Failing electronic library services A scholar’s tale* Jan Engh Bibliotek for humaniora og samfunnsvitenskap, University of Oslo, Norway Electronic services providing text and bibliographic information have become a necessary part of today’s research libraries. They are, however, surprisingly unreliable and unfunctional, and when they are combined locally, the deficiencies may get disastrous. However, these disadvantages have not received much attention, perhaps because each of them appears to be utterly trivial. It is time to focus on these drawbacks and the lack of attention that they have received from those responsible for their implementation and management in the libraries. Keywords: electronic library resources, texts, bibliographies, user experience, reliability, management, economy To the Norwegian linguist Tor Åfarli, internationally recognised, however far from recognised by electronic library resources Introduction Electronic services are a necessary part of any modern research library and great efforts are spent introducing increasingly more of them. Unfortunately, less efforts are spent on monitoring that the acquired services work properly in practice. This can be regarded as a technical problem only. It has also a social aspect, though, as it is conditioned by the relationship between the technicians introducing and managing the electronic services and their users, including library staff, especially those members who are, at the same time, active researchers. Further, malfunctioning of electronic services has economic aspects that have been largely ignored. Although basic logistics programs (on-line catalogues etc.) tend to work reasonably well it is my experience from a university library for the humanities and social sciences, that many other electronic library services are, in fact, infested with problems. The services in question are those representing full texts and providing bibliographic information1 among other things. It is true that there are fewer problems now than for instance ten years ago, at an early stage of electronic full text and bibliographic services. Still, there are problems and even if these may seem rather trivial and/or unimportant as seen one by one, they may, as a general phenomenon, cause frustration and even lead to complete resignation among users. Unfortunately, these problems are usually neglected by managers. Defects are fixed on an individual basis and no general solutions are attempted in cases were this would have been possible. Further, new versions or completely new software are purchased all the time on the basis of what are clearly insufficient requirements. * Thanks to Astrid Anderson, Svein Engelstad, Rolf Furuli, Gunn Haaland, Knut Hegna, and Diana Santos for providing examples or reading an earlier version of the manuscript. 1 Here, one may include even on-line lexicographic resources, e.g. Oxford English Dictionary. 1 Various kinds of defects are involved, relating to performance, conceptual quality, and local maladaptations. Or, to put it differently: reliability and functionality. In the following, these issues will be looked into. The perspective will be the one of a scholar and user who is supposed to utilise these services personally in his/her own research. On the other hand, aspects of electronic resources concerning the content of the services in a strict sense will not be discussed here: The question of identity between the printed and the electronic versions of journals that appear in the two modes,2 as well as the subject of data quality – from blurred illustrations and graphs3 to defects such as errors in titles and other bibliographic data, including incorrect representation of characters from outside the A – Z area on the screen and in outprints.4 In this article, I shall focus on how the content is transmitted, whether it can be used as intended and/or announced, and if it can be used the way a qualified user should expect. Additionally, aspects of the commercial side of electronic services will be discussed, as well as their management in a research library. Research Strangely enough, actual reliability and functionality for the individual user seem to have received little attention in specialist literature as opposed to extrinsic aspects of electronic library services: In fact, from Young (1997) to Evalued, passing other projects such as those reviewed by Poll (2005) and Sheeja (2010), the focus has been on organisational studies and sociological surveys. This applies even for titles that convey the suggestion of approaching the everyday search practice of scholars, such as Bar-Ilan and Gutman (2005), Ellis and Oldman (2005), and even Kiran and Singh (2008) and Buchanan and Salako (2009). What all analyses and surveys seem to take for granted is that the electronic services work as intended. The services can always be accessed and the functionality is perfect. One can always search and get what one wants. And, to the extent that the users are asked directly, it is all about whether they are content or not.5 Leaving aside a few quotations of user remarks, this is as close as one gets to the very personal experience of the users. Hardly ever details about how electronic library services actually work or ought to work in the everyday activities of scholars are accounted for. This impression is supported by this author’s inability to find articles, books etc. on the subject by means of electronic library resources available to him. However, as in other empirically based scholarly activities it is difficult to conclude that a given phenomenon does not exist. This even applies to what constitutes the object of the inquiry reported here. 2 For instance, the fact that the printed original of Clinical linguistics and phonetics 1988, 2/2 has 10 titles while the digitalised version contains only 3. 3 Cf. Engh, Haaland, and Rangnes [in progress]. 4 The “Helge Lødrup / Helge Lodrup syndrome”. These errors occur as a result of human decision, “simplification”, pure spelling error, or codepage confusion. Especially in the English speaking world, there is a tendency among technicians to regard letters such as à, ñ, and ø as “accented characters”, which may be represented roughly as non-accented characters. This is totally unacceptable and incomprehensible to any native users. As far as electronic texts are concerned, however, it is only fair to say that the situation has become much better after the generalisation of the PDF format. Cf. the exclusion of non English articles and the defective handling of non A – Z characters in English electronic journal articles offered by EBSCO at the turn of the century. 5 Cf. also the popular LIBQUAL survey concept, which otherwise is marked by an amazing lack of validity for linguistic and cultural reasons, at least in its Norwegian version. 2 Methodological remark The present article is based on my personal experience as a subject specialist (within the areas of linguistics and computation for the humanities) - from a regular day-to-day use of electronic library services in my capacity of active scholar and, at the same time, advisor to other researchers and students.6 In periods, electronic resources have been intensively tested in connection with separate linguistic research projects.7 Many of the specific cases referred to below occurred several years ago. Yet, they have been included since they are typical of the errors that still occur – which was one of the reasons why I started to write this article in the first place. No further extensive empirical research is necessary to prove the continued relevance of these errors, as anyone working in the field of humanities encounters similar cases almost on a daily basis - at least those of us based in a small country outside the English-speaking community and at a considerable distance from the publishers or agents.8 Essentially, this is a case study based on samples. As such, the result cannot be disproved. Its validity, however, depends on the experience of other scholars at the same institution, its interest in a wider context on the extent that scholars elsewhere experience similar problems. One caveat: Although the defects mentioned can be documented they cannot, in principle, be reproduced, since they have been properly reported and should have been subsequently fixed. Performance quality Do the electronic services function in the way intended or as announced? Do they work out in practice? And, if not, where can the cause be found? Access problems are by far more common in the case of electronic journals and books. So, I shall concentrate on this type of services in this section. The domain of the inquiry will mostly be services pertaining to linguistics and the humanities in general. On the provider’s side Many of the frequent failures to access the electronic journals or bibliographies are brought about by factors on the provider’s (publisher’s or agent’s) side. Basically, there are four possible causes: Failure to implement the subscription,9 insufficient computer systems, bad communication, and what appears to be singular technical errors, for instance alleged damaged files. Surprisingly often, the provider simply has failed to open the access as stipulated in the subscription contract, or the access has been switched off for one reason or another. 6 Working within the German “Fachreferent” tradition of academic librarianship well-known from the European continent. 7 Accounted for in Engh 2008, 2009, and [forthcoming]. 8 However, no further attempt will be made to relate the observations below to a geographical – or rather a centre/periphery dimension. Whether the situation described differs radically from the one of a research library localised in all respects close to the providers of electronic services has to be addressed elsewhere. 9 In fact, there is even the possibility of the library catalogue record announcing a non-existing/obsolete access. This lack of library quality, however, is outside of the scope of the present study. 3 As a consequence, the user is asked to log in and/or pay a per view fee. This is, at best, the direct result of mismanagement which is also indirectly involved when the computer fails to open access because of inadequate capacity: Apparently, the provider’s computer becomes unable to handle further requests as the number of concurrent processes gets too high, thus generating an error message, “We are currently unable to deliver the requested page due to peak demand. Please wait. Normal service will resume shortly”,10 or simply causing timeout at the user’s end. Bad communication has the same negative effect.11 In such cases, it is fairly easy to understand where the cause of the problem is localised. Other cases are less transparent in this respect. More or less intelligible error messages can be displayed on the screen of the user. Sometimes even the user’s own PC12 stalls as a result. One example: When this author tried to access the article Marquardt, Jacks, and Davis (2004) on 28 January 2005 Marquardt, Thomas, Jacks, Adam, and Davis, Barbara: 2004, ‘Token-to-token variability in developmental apraxia of speech: three longitudinal case studies’ Clinical linguistics and phonetics 18/2, 127-144 he was confronted with the message “The file is damaged and cannot be repaired”. This was also the case after a look-up of Gibbon (2004): Gibbon, Fiona E.: 2004, ‘Abnormal patterns of tongue-palate contact in the speech of individuals with cleft palate’. Clinical linguistics and phonetics 18/4–5, 285–311 Repeated attempts to look up these two articles on 31 January 2005 resulted in the same error message over and over again – before the PC stalled each time. Finally later on the same day, the last article appeared after a five minutes’ delay. Trying to look up this article 9 November 2009 resulted in the following error message: In order to access this article you must be a subscriber or you need to purchase the article. Please sign in to access your subscribed content, or select a purchase option. This case is typical also because it shows that, despite improvements, the problem is still there. To keep the record straight, my library had/has paid the subscription of the electronic version of the journal in question, and I tried to look up the article from a PC on the library’s premises. In 2005, the cause of the malfunctioning was not easy to find. Later, it became all too clear. There are, however, many cases where it may be almost impossible for the average user to determine what may be the cause of the problem encountered. And problems are, indeed, encountered. To get an idea about the frequency of failing access to electronic journals, an inquiry into the accessibility of electronic journals was carried out in 2004. During a one week test period (4 October to 11 October), I tried to look up articles in all of the 41 electronic journals pertaining to the discipline of linguistics and available at the library at that time. There were repeatedly problems 10 Error message from informaworld Taylor & Francis Journals Complete at a request for an issue of The new review of academic librarianship 12 April 2010 at 13:40 Central European Time. 11 It is still difficult to access bibliographic databases abroad after 14 CET, i.e. when the American East Coast started a new day. 12 During the last decade, the author has had at his disposal a series of state-of-the-art PCs, all of them perfectly capable of handling any relevant scholarly task. 4 accessing 9 of them. Additionally, the current issue of one journal was not accessible despite the contract. In the case of a couple of the nine defective journals, the probable cause was that the provider did not open the access (to the correct IP interval) despite contract and payment. One example: It took almost one week to re-establish the access to Linguistic variation yearbook. Failing access in spite of subscription was reported 4 October 2004. The access was re-established by the publisher according to contract - after two reminders - 11 October. In the remaining cases, bad communication or lack of computer capacity on the provider side were the probable causes. Thus, it would be possible to get access to the journals some other day – or even later in the evening that very same day. However, not always when desired or needed. This pattern repeated itself when I made daily attempts to download articles from the last issues of important linguistic journals such as Linguistics, Studies in Language, and Syntax in 2010.13 Were all these defects due to conditions on the provider side? Probably yes, although in other cases this may be a subject of discussion, for instance when an electronic journal can be accessed only by means of a VPN client and not via the library’s VPN portal. However, it is not of my business as a user to have any opinion on the issue. On the user’s side Clearly on the user’s side, though, (versions of) application programs may not work well together. This is the case of Adobe Acrobat and Internet Explorer and even of the plug-in version of Adobe Acrobat and certain Firefox versions. Some intermediary program or function may also cause problems. One example: In the Metalib search engine, there was recently a conflict between the portal and Internet Explorer version 8 concerning full text electronic journals: The user was supposed to look up the electronic journal in the union catalogue or in one of the other databases providing a link, then to click the link server button, “SFX”. Subsequently, one was supposed to click one’s way to the electronic article requested. So far, so good. However, if one wanted to look up a second article in what seemed to be the normal way, returning to the database and clicking the SFX button, while keeping the first article in its separate window on the screen, one would not advance any further. Instead, the SFX menu window would display the following simple but uninformative message: “Error on page”. In order to get around this problem, the user had to close the first window. This was far from obvious, and, of course, utterly unpractical for a user who needed to compare the two articles in one way or another – a rather normal need for most scholars. As an alternative, one not so obvious solution was to quit Explorer and use Firefox instead – but, at that time, it had to be versions 3.5.2 or 3.5.3. Problems of this kind are particularly irritating when they prevent making printouts. During the early autumn of 2009, for instance, it turned out to be impossible to print articles from the medical journal Lupus from a staff PC at the library. One makeshift solution was to mark the text of the PDF file on the screen and copy it to a Word file, created for this purpose only. Then this ad hoc Word file could be printed. (With a number of possible defects. See below.) Not exactly an elegant or efficient solution – and certainly not one according to the terms of the contract. 13 20 and 21 April and 30 April 2010. 5 So much for conflicting application programs. In other cases, one may even suspect that different versions of one single program, e.g. the full Adobe program used for creating the PDF file or the Adobe reader, may represent a problem. Finally, it goes without saying that the memory and the speed of the user’s PC are critical, as the machine readable information tends to become ever more extensive. During the first decade of the new century, at least my experience from an average research library shows that the renewal cycle of the PCs for public use did not keep up the same pace as the need for ever more computer power to display the articles of scientific journals. During the final years of the previous generation of user PCs, it was impossible to display – and make print-outs of – any electronic article except for the very short ones. The situation was not significantly more favourable at the faculty, where the staff’s PCs represented a broad array with varied characteristics as far as memory and speed were concerned. In practice, the problem of too small and too slow PCs is likely to repeat itself cyclically both on library premises and at the faculty, given the normal phases of renewal and updating of PCs and other electronic hardware. This is probably also the case with individual users off campus, having access through their own private PCs as remote users. Summing up The user is faced with a considerable number of problems as far as access to electronic library services are concerned, especially those providing texts. In general, these problems are of a trivial nature. They may, nevertheless, have serious consequences for the user. After all, fundamental aspects of electronic library services are concerned, such as access, screen display, and printing. In any case, the result is that the user can never be sure that (s)he will be able to use the electronic resources for which the library has a valid subscription. Conceptual quality Are the programs and the functions well conceived? Character encoding is essential, and we are, once more, talking about basic functions such as searching and copying. However, let me turn to the simple representation of a text before I proceed. Looking up the text Even when the access works according to the contract,14 viewing the text may be tricky. For instance, the NetLibrary instructions state that you just click on ‘View this book’ in order to look up a book. However, for many books one has to open one page in PDF format at a time. To see the next one, it is necessary to return to the appropriate NetLibrary page and click on ‘Next’ in order to see the next page. This is hardly functional, especially not when an entire book is concerned. However, few scholars would like to read entire books - fiction or monographs - on the screen. In the humanities, even articles tend to be rather extensive. So, reading the full 14 And nobody else is browsing the electronic book at the same time in the case of one loan at a time type of electronic subscriptions. 6 articles on the screen is usually not an option. In my view, searchability is, by far, the most important quality of electronic journals, ahead of copying for quotation. For this reason, deficient search facilities cannot be tolerated. Searchability Although non-English characters may be correctly represented on the screen and in outprints there are, in general, problems searching for anything else than A-Z, from Greek or Hebrew characters to accented Latin characters, not to mention the Norwegian (or Danish) characters Æ, Ø, and Å.15 For anyone with a scientific interest in the Norwegian language, it is natural to look up the name “Åfarli”. Tor Åfarli (1955-) is a renowned Norwegian linguist and a prolific scientific author. However, searching for “Åfarli” produces no hits. Searching for “farli” instead of “Åfarli”, however, makes you end up with a lot of hits, many of them “Åfarli”. It is, by the way, impossible to get around the character problem by copying for instance a Hebrew or IPA16 character from the article (or some other text format) to the search field, as any subsequent search does not work out. No normal linguist will have the competence/insight necessary to think of any more sophisticated way of searching. Copying text Copying text to other formats is equally a problem. Now, one may always object that this was never a requirement, despite all talk of compatibility between programs. Neither do the software producers promise any function of copying directly between their respective distinct formats. Yet, copying is one of the central assets of electronic texts, and definitely one of the reasons why they come in handy for scholars. That is: in theory. Depending on the purpose, copying means two things in our context: Copying from electronic journals to a database for further research use or quoting, i.e. copying to a manuscript (report, article, monograph etc.). In the following, only the latter will be looked into. It is a nuisance not to be able to copy margins, tabs etc. For instance copying from an electronic article in PDF format to a manuscript in Word format, using the special PDF copy function, inevitably leads to a collapse of the original format. Cf. this excerpt from Gavruseva (2000):17 5.1. Case and agreement relations in Germanic possessives The Germanic languages share two characteristics that distinguish them from the [+extraction] languages examined in sections 3 and 4. One characteristic is that some prenominal possessors are overtly marked for Genitive case by means of the ’s affix. The other characteristic is that they do not display ‘possessor agreement’ of the Hungarian type.13 In fact, the agreement relations in Germanic possessives are quite different. Consider first the data set in (27) that illustrates the Saxon genitive construction. Notice that in Norwegian, only pronominal possessors are allowed to bear the 15 I.e. ‘AE ligature’, ‘O slash’, and ‘A overcircle’. International Phonetic Alphabet. 17 This an the subsequent examples quoted in this paragraph result from copies made 19 April 2010 – with one notable exception. (See below.) 16 7 ’s affix: 13 An anonymous reviewer points out that the ’S affix in the data set in (27) below could be analyzed as a person marker due to its compatibility with a restricted set of possessors (proper names, third person singular or plural possessors, etc.) and the possible origins of ’s in the third person possessive pronoun (at least, in languages like Dutch). In this article, I follow an extensive body of literature that analyzes ’s as a Genitive case marker. I would like to note, however, that my arguments hold regardless of what analysis the ’s affix receives. [s. 761] (27) a. a. c. d. e. f. Peter’s book (English) bok-en hans (Norwegian, from Taraldsen, 1990) book-the his ‘his book’ Fyrirlestur Peturs (Icelandic, from Sigurðsson, 1993) lecture Peter’s ‘Peter’s lecture’ Johans bok-en (N. Swedish, from Svenonius, 1993) Johan’s book-the ‘Johan’s book’ Jans auto (Dutch, from Corver, 1990) ‘Jan’s car’ Peters Hunde (German, from Webelhuth, 1992) ‘Peter’s dogs’ (Gavruseva 2000, 760-761) which results in 5.1. Case and agreement relations in Germanic possessives The Germanic languages share two characteristics that distinguish them from the [+extraction] languages examined in sections 3 and 4. One characteristic is that some prenominal possessors are overtly marked for Genitive case by means of the ’s affix. The other characteristic is that they do not display ‘possessor agreement’ of the Hungarian type.13 In fact, the agreement relations in Germanic possessives are quite different. Consider first the data set in (27) that illustrates the Saxon genitive construction. Notice that in Norwegian, only pronominal possessors are allowed to bear the ’s affix: An anonymous reviewer points out that the ’S affix in the data set in (27) below could be analyzed as a person marker due to its compatibility with a restricted set of possessors (proper names, third person singular or plural possessors, etc.) and the possible origins of ’s in the third person possessive pro( 27) a. b. C. d. e. f. E. Gavruseva I Lingua 110 (2000) 743-772 Peter’s book (English) bok-en hans (Norwegian, from Taraldsen, 1990) book-the his ‘his book’ Fyrirlestur Peturs (Icelandic, from SigurBsson, 1993) lecture Peter’s ‘Peter’s lecture’ Johans bok-en (N. Swedish, from Svenonius, 1993) Johan’s book-the ‘Johan’s book’ Jans auto (Dutch, from Corver, 1990) ‘Jan’s car’ 8 Peters Hunde (German, from Webelhuth, 1992) ‘Peter’s dogs’ Of course, not only traditional formatting such as tabs, margins etc. are ignored. Graphs and diagrams are misrepresented as well. If, for instance, one tries to copy the following analysis tree from PDF to Word: Agr/DP (order of heads and specifiers irrelevant) (25) Agr/D’ Possessork Agr/D [+person] [+number] [+case] tk Possessum NP Spec N j Tj (Gavruseva 2000, 757) the result is (25) Agr/DP (order of heads and specifiers irrelevant) Possessor, AgrP’ AgrP [+person] >\ [+number] Spec N [+case] tk Possessum , G Imploding formats and graphs is a defect that most scholars can tolerate. After all, we are talking about different formats, not only different brand names. A systematic defect of greater importance when copying to a format outside the electronic journal is the garbling up of non Latin characters. Instead of the original characters, one gets nonsensical sequences of Latin characters and special characters. Cf. this excerpt from page 396 of Poirier (2004) after being copied to a Word file: “Podella writes of the [yqr in Ezekiel 1, “[yqr sonst nur noch im priesterschriftlichen Schöpfungsbericht als Bezeichnung des Firmaments, das den Himmelsozean abtrennt, an dem sich die ‘Leuchten’ des Himmels befinden, und als Bereich der Vögel, vgl. Gen 1,6.7.8.14.15.17.20; mehrfach in Ez 1:22.23.25.26 + Ez 10,3; den ‘Glanz des Firmaments’ nennt Dan 12,3. Ps 150,1 kennt in synonymem Parallelismus neben ‘in seinem Heiligtum’ (wvdqb) die Wendung “in der Feste seiner Macht” (wz[ [yqrb). Ist das eine Anspielung auf das himmlische Heiligtum? Und schließlich kennt Ps 19,2 die [yqr als Parallelbegriff zu ‘Himmel’ und als Ort, wo die Sonne ihr ‘Zelt’ hat, vgl. Hossfeld/Zenger, NEB 29, 133f; vgl. auch Keel, Jahwe-Visionen, 250ff.” (ibid., 201 n. 183). See also Morray-Jones’s discussion of the “celestial pavement” (A Transparent Illusion, 96-100).” 9 Here, [yqr wvdqb wz[ [yqrb is the rendering of the one of the one of רקיע בקדשׁו ברקיע עזו etc. Neither are phonetic characters rendered correctly. For instance, one paragraph on page 289 of Gibbon (2004) looks like this after being copied: EPG is a particularly valuable technique for investigating cleft palate speech because it provides information about a wide range of frequent errors. Errors in cleft palate speech most frequently affect sibilant fricative and affricate targets (/s/, /z/, /b/, / /, /tb/, / /), followed by plosive targets (/t/, /d/, /k/, /g/), then approximants (/l/, /r/, /j/, /w/) and finally nasals (/n/, /n/) and vowels (Spriestersbach, Darley and Rouse, 1956; Van Demark, Morris and Vandehaar, 1979). where /s/, /z/, /b/, / /, /tb/, / / and later /n/, /n/ are the renderings of /s/, /z/, /ʃ/, /Ʒ/, /tʃ/, /ʤ/, /n/, /ŋ/ respectively, and a table containing, among other things IPA fonts on page 488 of van Alphen and Smits (2004) reappears as:18 NW+comp bark daart bluim draan /b>rk/ /da7rt/ /bl!ym/ /dra7n/ bech dest bluk droost /bew/ /dest/ /bldk/ /dro7st/ biek dint brijs dwaalf /bi7k/ /dint/ /breis/ /dua7lf/ bijn doon broef dwijg /bein/ /do7n/ /bru7f/ /dueiw/ The original table looks like this: NW+comp bark /bɑrk/ bech /bɛ/ biek /biːk/ bijn /bɛɪn/ daart /daːrt/ dest /dɛst/ dint /dɪnt/ doon /doːn/ bluim /blɶym/ bluk /blʉk/ brijs /brɛɪs/ broef /bruːf/ draan /draːn/ droost /droːst/ dwaalf /dʋaːlf/ dwijg /dʋɛɪ Here, we are talking about systematic defects, in a number of electronic journals from different providers.19 This means that the very functions that constitute the basis for our need of electronic versions of scientific journals turn out to be deficient. But there is more. Not being able to copy characters outside the A – Z range is serious. Still, attempts at copying single words 18 Copied some time in 2009. Unfortunately, the very issue containing this article was defect 19 April 2010. Just for the record: Defect copying has no connection at all to installed fonts of Word etc. Further, exporting the text from the full Acrobat program into Word format is no viable solution, since the result is a copy which cannot itself serve as the source of screen copies. 19 10 or phrases usually results in easily detectable errors. However, even less easily detectable errors may be provoked by attempts at copying between different formats and programs. One such error is the omission of words at unpredictable places in the area marked when the word contains one single “Å”, a perfectly normal Norwegian character. This creates a critical situation when one intends to copy a paragraph relating to Åfarli, for instance. One example drawn from The Journal of Comparative Germanic Linguistics: Haider (1997). Using the PDF copy function to copy the following paragraph (page 18), According to Åfarli, the incorporated version is confined to participles in the analytic passive construction, as the contrast between (i) and (ii) illustrates. Herslund points out that prefixed particles are lexically conditioned and syntactically unproductive. the result is “According to farli, the incorporated version is confined to participles in the analytic passive construction, as the contrast between (i) and (ii) illustrates. Herslund points out that prefixed and syntactically unproductive.” For the keen observer, the sequence “particles are lexically conditioned” of the last line is missing, and “Åfarli” has lost his/its “Å”! Oddly enough, there seems to be some kind of relationship between the “Å” and the sequence of words missing somewhere else in the paragraph.). Now, this is what happens as long as the user does not have one of the last updates of Adobe’s Acrobate reader.20 With version 9.0 installed, for instance, this is the resulting copy: According toA° farli, the incorporated version is confined to participles in the analytic passive construction, as the contrast between (i) and (ii) illustrates. Herslund points out that prefixed particles are lexically conditioned and syntactically unproductive. This time, the entire text has been copied. However, the uppercase “Å” gets transformed to an uppercase “A” with no prior space and with a trailing circle plus the insertion of a new paragraph. This problem does not disappear when one moves from the free reader to the purchased (and presumably complete/extensive) version of the Adobe program. But this is beside the point anyway, since the users – at the faculty, at home etc. – have quite a number of different versions of the program. In fact, one cannot expect every student and every professor to have the most recent version of all programs. This is, in turn, related to the fact that their PCs vary considerably in terms of capacity, rapidity etc. Summing up Programs have problems communicating despite all talk about compatibility. The technical reasons for this may be clear and declared: The text formatting is different and PDF uses UTF-8, while Microsoft Word uses a proprietary Windows-based character set. Still, this represents a general - mostly ignored - problem. In fact, one may wonder why this problem hasn’t been fixed long time ago. However, the users have every reason to expect a seamless transition between them. The lack of it is a true annoyance and may 20 In this case, version 5.0. 11 even result in loss of material/content and errors for those unsuspecting and/or in a hurry. In the last case above, even being aware that defects of this kind may happen requires an extraordinary level of attention from the user – or a stronger competence in basic computing that can be expected from normal academic users. In the extreme, lack of conceptual quality will harm the aspects of higher education and research that electronic library services are supposed to facilitate. The search results may be distorted, and, in their daily use, the library services may appear to be unpractical. Local maladaptations Deficiencies may occur due to local administration of the respective purchased programs or functions as well. Both electronic full text databases and bibliographies may be involved. One outstanding example is the ARTIKLER option of the search field in the very centre of the homepages of all the library branches at the University of Oslo. This is a federated search in a selection of resources via Metalib. In order to look up every record related to a linguist or a writer of fiction, for instance, one is compelled to search a bundle of preselected electronic resources labelled “humanities”:21 Humanities Index, JSTOR, MLA international bibliography (PQ), Periodicals index online, and Philosopher's Index (Ovid). These are truly odd bedfellows from several points of view. On the one hand, a medley of bibliographies and full text scholarly journals. On the other hand, radically different disciplines such as linguistics, literature, philosophy, and - as will be shown in an instance - history of art.22 As a consequence, no result satisfying the needs of higher education or research is obtained. Two examples will make this clear: once again the linguist Tor Åfarli and Thomas Hardy, the British author. As already indicated above, “Å” has the potential of messing up the result of even the simplest program or function. Thus, searching for “Åfarli” in ARTIKLER will give as its result23 a small but awkward number of hits - 13 in all: 1 2 Afarli, Tor A., reviewer ýfarli, Tor A. 3 ýfarli, Tor A. A promotion analysis of restrictive relative clauses 1994 4 Haegeman, Liliane Review of Tor A. ýfarli, “The Syntax of Norwegian Passive Constructions” 1994 Back Matter Publications Received “The Syntax of Norwegian Passive Constructions”, by Tor A. ýfarli (Book Review) 1993 1993 1993 5 6 7 Lýdrup, Helge Coordination (Book Review) 2000 A Note on the Syntax of Adverbial Phrases 1995 21 Humanities Index Periodicals index online Periodicals index online Periodicals index online JSTOR JSTOR Periodicals index online “Humaniora” in Norwegian. Of course, the user has the possibility to search a variety of relevant resources as well, which are listed “deeper” inside the library pages. However, ARTIKLER is the one that the library promotes for any scholarly search. 23 Via Firefox. 22 12 8 ýfarli, Tor A. Passive in Norwegian and in English 1989 9 ýfarli, Tor Nonsubject Pro-Drop in Norwegian 1987 10 ýfarli, Tor Norwegian Verb Particle Constructions as Causative Constructions 1985 11 Deborah Marrow 1979 12 13 Malcolm Waddingham Keith Andrews Maria de’ Medici and the Decoration of the Luxembourg Palace Elsheimer Revised The Elsheimer Inventory and Other Documents Periodicals index online Periodicals index online Periodicals index online JSTOR 1972 1972 JSTOR JSTOR The weird character returned for “Å”, “ý”, is quite funny, but this error has no importance.24 What is important, though, is the fact that there are only 13 hits. 13 extraordinary hits, indeed: 1 is a review written by Åfarli. 2, 3, 8, 9, 10 are articles written by Åfarli 4, 7 are reviews of two of Åfarli’s works (written by others) So far, so good. However, the following hits 5 6 is an advertisement for Åfarli’s monograph The Syntax of Norwegian passive constructions by John Benjamins is an item in a list of books received, Language 69/2, 441-448 are hardly enlightening and of marginal relevance. But the worse is still to come: 11 is one token of the sequence “afarli”25 in a quotation in Italian in an article on the history of art: 3. Med., 4639, not paginated, 31st December 1623, Picchena to Gondi. ... Quanto ai quadri di Pittura che la Regina vorrebbe per adornare le gallerie del suo Palazzo, V.S. potra dirle che Lor Altezze n 'hanno gia dato l'ordine, et procureranno che ella sia servita nel miglior modo che sia possibile, sebene questi Pittori, massime i principali, sogliono per ordinario esser molto lunghi, et non basta il sollecitarli, durandosifatica afarli uscir di passo. Perb se questo servizio andasse un poco alla lunga, non bisogna maravig- liarsene, et io in particolare non mancherb di dar lor fretta, accib S. M..t sia servita presto.... 4. Med., 4 From Marrow, Deborah: 1979, “Maria de’ Medici and the Decoration of the Luxembourg Palace”. The Burlington Magazine 121/921, 783-79126 12 refers to two token of the sequence “farli” in a quotation in Italian in another article on the history of art: 24 Remark that “ý” is also used to represent “ø”, ‘o slash’. “Lýdrup” instead of “Lødrup”. Additionally, even “A” is used instead of “Å”, cf. “Afarli”. 25 “farli” is a form of the Italian verb FARE ‘do’ plus an enclitic pronoun. “a” is a preposition. 26 Even including “farli” as a constituent of the compound verb SATISFARE ’satisfy’. 13 B.mo Padre. Carl' Antonia figlia del Conte Vilglierm di Stuard Scozzese devotissima serva de V. B.ne ch' in Roma, e venuta solamente per fuggire I'heresie di quel paese, e seguire lafede catolica, mentr' era vedova, non sendo socorsa d'alcuni aiuti, ch' aveva al tempo di Clemente VIII, per alimentare se stessa, et unfigliolino, e stata costretta difare qualche debito, per li quali e, molestata, et ogni giorno si rilassano contra di lei mandati, e perche non si trova commodita di denari, per satisfarli, accio non gli si levi, quanto ha in casa et cosi resti, in tutto nuda, suplica V. S.ta afarli gratia di comandare a Monsignor Governatore, nel cui tribunale, e stato ora relassato un mandato che li ceda competente dilatione, di pagare una certa somma ogni mese, altrimente sara forzata di ritornarsene alla patria sua, che il tutto ricevera, per somma gratia di V. B.a Quam Deus, Carl' Antonia d' Stuart From Waddingham, Malcolm: 1972, “Elsheimer Revised Author(s)”. The Burlington Magazine 114/ 834, 600-611 13 refers to a fourth token of the sequence “farli” in a quotation in Italian in an article on the history of art: This token is identical to the one of 12. The very same Italian manuscript is quoted in Andrews, Keith: 1972, “The Elsheimer inventory and other documents”. The Burlington Magazine 114/ 834, 595-607 This means that the result of a search for “Åfarli” produces as a result 2 reviews of Åfarli’s works and 6 references to his own works, which is rather unsatisfactory, given in the fact that Åfarli himself has included 175 titles in his own personal bibliography (November 2009).27 On top of it all, the remaining 5 hits are either slightly irrelevant (5, 6) or bizarre (11, 12, 13). (Apparently, somewhere in the system a code page error deletes the “Å” or converts it into an “a”, and a search for the distorted name of the linguist, “farli” returns as a result 4 tokens of the Italian inflicted form of the verb FARE plus an enclitic pronoun in articles from English arts history journals reproduced by JSTOR. Clearly, “Å” represents a problem. However, this special Norwegian character is not the only cause of the unsuccessful search. Unfortunately, the same lack of sensible results ensues from a search where the search condition contains only characters from the A – Z range and where the subject is of general international interest. One such example is a search for the British novelist and poet Thomas Hardy (1840-1928). The result is 117 hits. Their nature or quality may be more relevant. What is important, however, is that they are much fewer than one should have expected: A similar search in Literature online (‘Criticism & Reference’ > ‘criticism’ > ‘subject’ = ‘Hardy, Thomas’, October 2009) results in 8704 hits. So, even in the case of central British authors with names well inside the A-Z range the result represents a tiny fraction of what one will get from one single search in one of the constituent bibliographies. Confronted with such a catastrophic result, any user will drop the ARTIKLER function or worse. 27 http://wo.uio.no/as/WebObjects/frida.woa/wa/fres?action=sok&etternavn=%C3%85farli&fornavn=tor&er Norsk=1&erNordisk=1&erNasjonalt=1&erInternasjonalt=1&erUkjent=1&visParametre=1&bs=50 [November 2009] 14 Discussion Using electronic full text and bibliographic library services, one may experience repeated access failures, failures to print, and copying problems, lack of search possibilities or absurd search results - at least within the domain of linguistics and the humanities in general. From a management point of view, each instance of such defects may be dismissed as singular, or in the case of conceptual defects, as irrelevant to library services. Inevitably, users will look at this in a different way. The importance of the defects increases because of their quantity and the users get the impression of negligence and general lack of reliability on the part of the library. Persistence over the years emphasises this negative impression. On-line full text and bibliographic services have been around for more than a decade in their actual forms. Still, their reliability is poor and their functionality far from adequate, at least as far as services within linguistics and the humanities are concerned. How can this have continued for such a long time? Indirectly, one reason may be that, in many academic disciplines, for instance in the humanities, electronic journals and bibliographies are not indispensable. As opposed to what many seem to believe, they are, basically, just an extra asset. A complimentary explanation is that researchers may have made private/individual arrangements or local contracts at department or research group level doubling those of the research library.28 This may also be reflected by the fact that relatively many defects are reported by library staff members, not by ordinary users.29 From a direct point of view, the defective performance of these services has continued because of lack of managerial attention, which manifests itself in two ways: Lack of commercial attentiveness and technical management’s negligence of users experience, including the one of qualified members of the library staff. As distinct from what is common business practice, just the mere functioning of services is checked by the library – however, only at the point of reception. No systematic check of their continued functioning, their day-to-day availability, is performed,30 not even their continued functioning after contract or term renewal. It is simply presupposed that all acquired electronic resources work well and the users are expected to report possible errors on an individual basis. To judge from the diversity between defects experienced by library personnel and those reported by ordinary users, the inevitable result is that far from all cases are reported. Further, the scarce individual error reports are redirected to a librarian, who checks if the library has a subscription or not and eventually makes a complaint to the provider of the service in question. From one 28 In fact, many researchers have their own subscriptions of printed scientific journals. A phenomenon that strikes me as similar to the use or non use of other optional electronic functions. As a member of an advanced international research and development team in the 1980s, I was puzzled to see the users’ reactions to the earlier and not so intelligent software. Using word processing software, for instance, they easily lost confidence in spelling checkers completely when they saw that a screenful of text contained many “incorrect” or “unrecognised” words that were, in fact, correct. The users tended to turn the function off, never to turn it on again. This author’s experience as a tutor of electronic library resources is that the same tendency manifests itself as far as certain electronic library services are concerned. After just a few failures, the users simply do not use them again – without reporting any errors. 30 Unlike the traditional checking of every purchased book or journal issue at the point of reception and later at regular shelf reviews and in more singular occasions, for instance when complete volumes are selected for binding. 29 15 point of view, this is a rational way to proceed, but it functions on a case-to-case basis, and no statistics of the defects are established that could serve as the basis for general administrative action.31 In fact, the users may even have problems understanding where to report what, since they have every reason to be confused as to where their problem is localised in the chain of journal – publisher/agent – communication – library/local net own PC. As for conceptual quality, this is, primarily, not the responsibility of local electronic services’ management once a program has been acquired and put into production. Still, the latent partial lack of compatibility between existing programs and functions is of great importance for the users and should attract the library’s attention as such. Unfortunately, problems of this kind are never properly addressed. No systematic testing of functionality is carried out at any stage prior to acquisition, nor are functional defects systematically reported – unlike what is common in software industry. And, perhaps most important, no firm requirements as to true compatibility are set when electronic resources are acquired. What is the clear responsibility of local managers, though, are local solutions built on purchased services of various kinds. In the end, the result will constitute the product of its sub-systems’ defects and inherent weaknesses despite the quality of its components when used one by one. One prominent example is the ARTIKLER function described above. Lack of checking, testing, and adequate requirements all have to do with the libraries’ lack of professionalism as software customers. Apparently, this is a property that characterises the entire trade: Since the unreliability of electronic resources has turned out to be a persistent phenomenon it is obvious that the research libraries have not collectively exerted their power as customers. Another dimension of the lack of managerial attention is, ironically enough, the negligence of the qualified users, including in-house expertise where that can be found. In the case I am referring to, the experience of subject specialists is systematically left out of account. This holds for reliability, conceptual quality, as well as for locally concocted functions. The ARTIKLER function was introduced despite unison criticism and thorough documentation of its contra productive effects from the library’s own staff of subject specialists. Later on, reports of malfunctioning have been ignored and those raising the subject turned away.32 31 Here, it should be added that the users often have no obvious place to turn to report errors. In the University of Oslo Library case, for instance, one tiny button “Hjelp/tilbakemelding”, literally ‘help/feedback’, hidden in the list of providers and stock information in an intermediary screen is hard to find and clearly not enough in order to create an impression of readiness to help on the library side. Cf. for instance http://x-portsfx.uio.no/sfx_ubo?sid=BIBSYS&date=9999&sfx.ignore_date_threshold=1&rft.object_id=1109789665591 44 [19 April 2010] 32 Postlude: When, after one year, the ARTIKLER function was finally removed from the library home page in 2010, the motive was to copy the new homepage of another university library. Thus, the ARTIKLER function was finally substituted by – a pointer to Google Scholar. Without any previous consultation of the local scientific staff or scholars of the faculty. 16 What needs to be done? Obviously, a first measure33 would be to abstain from improvising local “solutions” such as ARTIKLER. Further, existing services have to be monitored systematically. Software for checking that the respective resources are fully accessible at any time should be implemented - for the benefit of the users, but also for obvious economy/audit reasons. Why should libraries pay for services not provided …? In this context, it is even appropriate to suggest that one ought to control all issues of an electronic journal at reception. This, however, will inevitably be more complicated than the traditional control of printed issues, since a scientific officer will have to check that the printed and electronic versions are identical as far as content is concerned, perhaps even the functionality of the electronic version. As far as further acquisitions are concerned, new and tougher purchase requirements must be set up, taking formal errors (e.g. failing to open access according to contract) and technical aspects (e.g. access capacity, virtual network software compatibility) into consideration. Additionally, requirements should be established for true compatibility between all new programs and their functions as far as conceptual quality is concerned and serious and decisive efforts should be made to harmonise partly incompatible functions in different existing programs. As a prerequisite, however, a change of attitudes is required among managers of electronic library services: Apart from adopting a less naive approach as far as the commercial aspects of electronic services are concerned, they have to be less driven by advertisements, fashions, and impressions from meeting their equals. Instead, they ought to take an interest in the users’ concrete experience with the electronic services and how they actually want to use them.34 Here, cooperation with the rest of the library staff is essential, especially with members who have full experience as students and researchers within the academic disciplines involved. Their insight in the multifaceted properties of the various academic disciplines provides the key to both the adaptation of electronic library services as well as to their continued control and future development. Bibliography Specialist literature Bar-Ilan, Judit and Tatyana Gutman: 2005, How do search engines respond to some nonEnglish queries? Journal of information science 31/1, 13-28 Buchanan, Steven and Adeola Salako: 2009, Evaluating the usability and usefulness of a digital library. Library review 58/9, 638-651 Available at www.emeraldinsight.com/0024-2535.htm [1 September 2010] Ellis, David and Hanna Oldman: 2005, The English literature researcher in the age of the Internet. Journal of information science 31/1, 29-36 33 As far as software is concerned. A precondition is, of course, that the research library’s hardware is conform to an adequate standard. The PCs, for instance, ought to have a capacity matching the technical requirements of the acquired software, or, to put it differently, it should be possible to run new software on all the PCs of the institution and on those of every user. 34 For obvious reasons, surveys of general levels of contentment are not sufficient. 17 Engh, Jan: 2008, Norwegian examples in international linguistics literature. An inventory of defective documentation. München: LINCOM Engh, Jan: 2009, Defective documentation. International linguistics and modern Norwegian. Folia linguistica 43/2, 269-310 Engh, Jan: [forthcoming], Refereed defects. Norwegian in international linguistics publications Engh, Jan, Gunn Haaland, and Kristin Rangnes: [in progress], Uklare illustrasjoner i digitale tidsskrifter [‘Indistinct illustrations in digitalised journals’] Evalued. Evaluating electronic library services. [S.a.] UCE library services. University of Central England, Birmingham UK. Available at http://www.evalued.bcu.ac.uk/summary_booklet.pdf [1 September 2010] Kaur, Kiran and Diljit Singh: 2008, Exploring user experience with digital library services. A focus group approach. In George Buchanan, Masood Masoodian, and Sally Jo Cunningham (eds.): Digital libraries. Universal and ubiquitous access to information. 11th international conference on Asian digital libraries, ICADL 2008, Bali, Indonesia, December 2-5, 2008. (Lecture notes in computer science 5362), 285293 Poll, Roswitha: 2005, Measuring the impact of new library services (IFLA Oslo 2005). Available at http://archive.ifla.org/IV/ifla71/papers/081e-Poll.pdf [1 September 2010] Sheeja, N.K.: 2010, Undergraduate student’s perceptions of digital library. A case study. The international information and library review 42/3, 149-153 Young, Peter R.: 1997, Measurement of electronic services in libraries. Statistics for the digital age. 63rd IFLA general conference – conference programme and proceedings – August 31-September 5, 1997. Available at http://archive.ifla.org/IV/ifla63/63youp.htm [1 September 2010] Examples van Alphen, Petra M. and Roel Smits: 2004, Acoustical and perceptual analysis of the voicing distinction in Dutch initial plosives: the role of prevoicing. Journal of phonetics 32/4, 455–491 Gavruseva, Elena: 2000, On the syntax of possessor extraction. Lingua 110, 743-772 Gibbon, Fiona E.: 2004, Abnormal patterns of tongue-palate contact in the speech of individuals with cleft palate. Clinical linguistics & phonetics, 18/4–5, 285–311 Haider, Hubert: 1997, Precedence among predicates. The Journal of comparative Germanic linguistics 1, 3–41 Poirier, John C.: 2004, The ouranology of the apocalypse of Abraham. Journal for the study of judaism in the Persian Hellenistic and Roman period 4, 391-408 18