Since 2007, Ethnologue’s three-letter codes for languages have had the status of an ISO standard for languages. This has considerably enhanced their status in linguistics, and some linguists now use these codes (which were primarily intended as unique identifiers for technical and industrial purposes) in their prose texts, as additional identifiers of the language(s) they are talking about. But at the recent PARADISEC conference in Melbourne, Stephen Morey, Mark W. Post and Victor A. Friedman launched a direct attack on them (Morey et al. 2013). It seems to me that this is a very useful step, because it may lead to a serious public debate of what it is that we can expect from a standard language catalogue such as Ethnologue. (Some discussion of the issues has been taking place in closed circles or small workshops, e.g. at the Leipzig language catalogue workshop in 2007, but this is the first highly critical voice at a major conference, as far as I know.)
Morey et al. have five main points of criticism, the first two of which could in principle be addressed by changing the procedures (and have actually been addressed by Glottolog):
(1) Three-letter codes are problematic, because quite a few of them are based on obsolete and offensive designations for languages and speech communities (e.g. [jnj] for Yemsa, based on pejorative “Janejero”).
(2) Administration of ISO 639-3 codes by SIL International is problematic because it is a missionary organization, which does not work with sufficient transparency and accountability.
(3) Permanent identification of a language is incompatible with the constantly changeable nature of human language.
(4) Languages and dialects cannot be readily distinguished by rigorous and practical methods, and inevitably, socio-political concerns often enter into the decisions taken by linguists.
(5) ISO 639-3 has the potential to be misunderstood and misused by governments and other decision-making bodies.
They also voice objections to Ethnologue’s attempt to classify all languages into subfamilies and families, but this is quite independent of ISO 639-3, so this issue should be kept separate from the language code issue.
I share their concern with issues (1)-(2) and (4)-(5), but point (3) is not a serious objection. Yes of course, languages change all the time, but it is also the case that we can identify them and write papers and books about them. We can also give different identifiers to different stages of a language. So this is not a reasonable objection.
The main question that I have is whether language identification should be a task for ISO, the International Organization for Standardization. As Morey et al. note, ISO provides standards to ensure that “materials, products, processes and services are fit for their purpose”. ISO is basically an organization or industry, not for science. Of course, scientists need to identify their phenomena and catalogue them, but normally it is scientific organizations that take on the task of standardizing names of entities, such as the International Mineralogical Association that administrates names of c. 6,500 minerals (roughly on the order of the number of languages), and it is the International Astronomical Union that takes care of naming issues with regard to heavenly bodies. The reason why ISO got involved in language name issues in the first place is of course the economic significance of translation and localization, which is far greater than the relevance of distant stars for businesses. But does this mean that someone needs ISO’s industry standard to identify little-known languages of small communities that are never or hardly used in writing and that are often in danger of extinction?
The inappropriateness of ISO as an organization for maintaining nomenclature standards for linguistics was brought home to me when in 2011 I organized a small workshop on language cataloguing in Leipzig. I was aware that ISO had created another standard called 639-6, which also contains codes for language families – obviously of great interest to linguists as well. I found out that 639-6 was entrusted to an organization called Geolang Ltd. I invited a representative of Geolang to our workshop, who turned out to have a lot of technical expertise, but zero knowledge of comparative linguistics. This seemed very odd to me – at least SIL International has demonstrated expertise in the study of languages!
What is the solution? Surely we do not want to do without any unique identifiers of languages. These are simply too useful for tagging books and articles in libraries, for maintaining catalogues of resources such as OLAC, for linking between encyclopedias like Wikipedia and other resources. So how do we address the issue of problems in distinguishing between dialects and languages? How do we address the problems of accountability and transparency? What about potential misuse?
It seems to me that the best solution is not to have one single place where language identification information is stored and maintained, but multiple places that link to each other. And in fact, de facto there are are already four authoritative places: In addition to Ethnologue, there is Wikipedia, MultiTree and Glottolog. These all have different philosophies, but they are partially interlinked: Ethnologue links to OLAC, which links to Glottolog; Wikipedia links to Ethnologue; and Glottolog links to everything else. So from Glottolog, one can easily go to the other places and check out what they have to say about the language that one is interested in at the moment.
Like Ethnologue, Glottolog (Nordhoff et al. 2013) has unique identifiers for languages, but these look less like abbreviations (they consist of four letters and four digits), so it is hopefully clear that their purpose is a purely technical one, to make it easy for machines to link one resource to another one. They should not appear in the body of a scientific text (at most in a footnote), which addresses problem (1) above. Glottolog aims to be fully transparent and accountable, addressing problem (2) above, though its degree of transparency is currently hampered by not fully adequate funding. In addition to languages, Glottolog also lists dialects, thus partly addressing issue (4). In the future, we may switch to a system where languages are recognized at multiple levels, so that everything is recognized as a language that anyone ever called a language. After all, it rarely matters to linguists whether what they are talking about is a language, a dialect or a close-knit family of languages. Thus, linguists may choose to work at the “languoid” level, ignoring the difference between languages and dialects. This addresses problems (4) and (5).
Some readers may object to Glottolog’s introduction of an additional code (the “Glottocode”), because it would be more practical if there were just a single code, and we could somehow work together to improve the current ISO 639-3. But I don’t see any single authority that could impose its views. We have no International Union of Linguists (CIPL is underfunded and many linguists have never heard of it), and in general, knowledge of and interest in languages is very decentralized, and different people will always have different views. So it seems better to accept the diversity of approaches, but to adopt Glottolog’s model of keeping multiple identifiers and linking to several other language catalogues. This does mean that a certain amount of work is duplicated, but it also means that there is less danger of a single dominant point of view eclipsing dissenting voices.
References
Morey, Stephen & Post, Mark W. & Friedman, Victor A. 2013. The language codes of ISO 639: a premature, ultimately unobtainable, and possibly damaging standardization. Handout of a talk given at the PARDISEC RRR Conference, December 2013.
Nordhoff, Sebastian & Hammarström, Harald & Forkel, Robert & Haspelmath, Martin (eds.) 2013.
Glottolog 2.1. Leipzig: Max Planck Institute for Evolutionary Anthropology.
OpenEdition suggests that you cite this post as follows:
Martin Haspelmath (December 4, 2013). Can language identity be standardized? On Morey et al.’s critique of ISO 639-3. Diversity Linguistics Comment. Retrieved October 4, 2024 from https://doi.org/10.58079/nsst
I agree with almost all of Martin’s post (thanks for writing it, Martin!), and am certainly hopeful that the current and inappropriately monolithic regime can be replaced by something more diversified, transparent and participatory, perhaps along the lines that MH suggests. This would probably address the majority of the problems we list in our paper, and would probably make most linguists comfortable with “standard” sets of labels for languages in general.
But although it may seem like a minor issue in the overall context, I’d like to briefly address MH’s point that our claim that “permanent identification of a language is incompatible with the constantly changeable nature of human language” is “not a serious objection” to the ISO 639-3 (or any other) language-coding regime. MH writes, “yes of course, languages change all the time, but it is also the case that we can identify them and write papers and books about them.”
Fine; but the fact that we refer to an entity in our linguistic and academic practice doesn’t necessarily make it real in the sense implied by particular types of label. We make the point in our paper that human languages are observably not bounded, static entities, but are instead open-ended, dynamic processes. Put differently, “languages” are methodological abstractions that are made over observable patterns of human behaviour, rather than the observable entities that an ISO-style naming practice implies.
As others have pointed out, the ISO is primarily oriented toward efficiency in business and administration; it develops unambiguous and explicitly permanent symbol-referent associations, for the purpose of reducing errors, waste, confusion, etc. If we look at the types of things ISO names, they include things like chemical compounds, weights, measurements, energy management and food storage practices – entities or processes which are either naturally-occurring or arbitrarily specifiable, but in all cases delineable and observable. If we look at the types of things ISO doesn’t address, they include human emotions, hairstyles, dance styles, genres of film and other cultural features. Now, it’s possible that there just hasn’t been enough demand yet, but it’s also possible that ISO will never touch these or any similar culturally-evolved processes or practices because of their inherent unboundedness – their amenability to imprecise, fuzzy and evolving natural-language reference, and their non-amenability to precise, stipulated, and permanent machine-language reference. Our claim, which we are hardly unique in making, is that human language, and human languages, observably fall into the latter group rather than the former. Is it any easier or more difficult to distinguish in a time-stable way between an “emo” and a “goth” hairstyle than it is to distinguish between dialects p and q of language x? Or between language x at time periods a and b? We’d say no – almost by definition, since the variables identified here represent arbitrarily-chosen points along evolving continua. Human language has developed various means of coping with a world of shifting referents – let the denotation of a symbol evolve together with the referent, or else develop a new symbol. In other words, be imprecise (at least, over time). Any ISO-style standardization regime, i.e. one which assumes a permanent symbol-referent association, will fail in this regard; this is certain. What worries us is that implying the opposite has the capacity to mislead and cause harm.
Now, we obviously need conventional labels of some sort for the languages that we write about. But whereas the current and imprecise naming practice among linguists more closely resembles natural-language naming practices, and responds to changes as they occur on the ground (for example, by adopting new labels when older labels become pejorative), and whereas we, as linguists, can always include disclaimers in grammars regarding what we perceive to be the proper relationship between a descriptive grammar and the language data, the ISO 639 regime, at least, doesn’t and probably will never do this. And while these sorts of issues can be made clear to a linguist, more or less, they aren’t ever going to be clear to a bureaucrat.
Over the last two months since the PARADISEC conference, I’ve grown hopeful that the novel term “languoid”, being more technical-sounding (or at least more obscure) than “language”, might help solve this problem – ironically, by making the category itself seem, at least, less transparent, and thus perhaps less useful to non-specialists who might otherwise have wanted to make use of an ISO-type labelling regime!
Mark:
Your remarks seem to me to assume that the role of ISO 639 is (or should be) one of scholarly documentation, and that anything coded by ISO must be done so in a way that would be consistent with correct, scholarly documentation. But that is not the role of ISO coding standards.
There might not be an objectively correct taxonification of human emotions, but if a need arose in information systems to convey certain categories of emotion in metadata protocols for purposes of interoperable information exchange, it might well become appropriate to have an ISO standard to encode those categories of emotion. That would not be—and should not be construed as—an assertion of a correct analysis of types of human behavior; it would merely be a technical means of representing certain categories of information in an inter-operable a way in order to facilitate whatever usage scenarios gave rise to the need for such representation in the first place.
There is no question that languages are dynamic, fuzzily-bounded entities. Even so, there is real need in information technologies to qualify information objects (or associated queries) to indicate associations with concepts such as “English” or “Dyirbal”. That necessarily requires assuming at some level well-defined notions of “English”, “Dyirbal”, etc. And, of course, there will always be possible counter-examples refuting the validity of assuming such notions to be absolute and well-defined. Nevertheless, the notions can maintain a very high level of usefulness for a large number of users. And that is what makes them appropriate for codification in an ISO standard.
In other words, it’s recognized that a standard like ISO 639 is not and can never be perfect, but it is practical, and practical is good enough. (That’s not meant to rule out possibilities for improvement in implementation of the standard.)
By the way, all the same considerations apply equally to coding of “characters” in ISO/IEC 10646 (ISO equivalent to Unicode), and similar debates between technologists and scholars have occurred in that domain as well.
Peter:
Thanks for these very helpful comments! In reply,
> Your remarks seem to me to assume that the role of ISO 639 is (or should be) one of scholarly documentation, and that anything coded by ISO must be done so in a way that would be consistent with correct, scholarly documentation. But that is not the role of ISO coding standards.
If that’s so, then I’m pleased to be corrected on this point! Unfortunately, however, many (most?) linguists and linguistics-oriented organizations (e.g. granting organizations such as ELDP) are operating in practice under the same misapprehension that I was under. If it is the case that ISO 639 has only the goal of being “practical” and “good enough” to manage information systems, and should not be expected to reflect scholarship in any substantial way, then linguists and linguistics-oriented organizations should simply ignore ISO 639, and/or produce a different regime more tailored to our needs and expectations (a point which I take MH’s post to also be arguing). Similarly, Wikipedia should cease the misleading practice of including ISO codes together with pages describing languages, since Wikipedia’s readers and contributors generally *do* believe that Wikipedia’s pages are or should be based upon “correct, scholarly documentation”. In this sense, the fault is not at all with ISO’s standards, but rather with those of us who have been misusing them; seemingly, most of us probably shouldn’t have been using them at all! I’d agree with that.
“Most of us probably shouldn’t have been using them at all!” That’s an excluded-middle fallacy. Most people _should_ be using them… when appropriate. (For most people, that’s most of the time.)
A group of politicians and historians could have lively debates about the borders of India in the course of which they might refer to many entities that should not be represented using the ISO 3166 identifier “IN”, but that’s a somewhat specialized context. If, on the other hand, they want to create a Web site in India, they should expect a domain name ending in “.in”.
There will certainly be situations in which it would not make sense for a linguist (or anthropologist or…) to rely (or rely solely) on an ISO 639 identifier to represent a particular speech variety.* But there will still be scenarios in which you absolutely should use it. For example, if you create content in (say) Dyirbal in an HTML or XML document, then you should use lang=”dbl” / xml-lang=”dbl”. If you’re using XML for some linguistic corpus and need to capture details not appropriately representable in ISO 639-3, then you could define a schema that includes some other attribute or element to capture such information using proprietary metadata elements defined within your project or some convention of metadata elements shared among a body of practitioners in your field. (You might still use xml-lang and supplement the other metadata, or dis-regard xml-lang. But you shouldn’t use xml-lang in a way not sanctioned by the XML spec — a bit more below.)
* ISO 639-6 was intended to provide a comprehensive coding of sub-language speech varieties, and from the start I considered the approach a bad idea. That particular standard will come up for a systematic review this year; you can recommend to the relevant committee within your country’s national standards body that it be withdrawn.
A bit more on xml-lang (and html lang): The XML spec references an Internet (IETF) specification known as BCP 47 (http://tools.ietf.org/html/bcp47). That spec defines a system of tags for representing languages and language varieties. It’s a very flexible scheme that can accommodate a wide range of language varieties, within the constraints of using ISO 639 as a starting point. For example, the tag “sl-rozaj-biske” can be used to represent the San Giorgio/Bila variant of the Resian dialect of Slovenian; “fr-1694acad” can be used to represent Early Modern French as codified in the 1694 edition of Dictionnaire de l’académie Françoise. Unlike ISO 639-3, BCP 47 allows users to register language variant subtags to accommodate a particular need, rather than attempting to create an a priori view of the world. Also, if needed, BCP 47 allows for private-use extensions that would allow you to represent whatever you want; e.g. “en-x-2ndlang-clftpal”. In addition, it allows for conventionalized extensions that could be devised by some body of users with common needs. For example, a group from the software industry has defined a “u” extension that is used in tailoring software processes for things like constructing date or time strings to match the expectations of a particular culture.
A quick comment on Wikipedia: I have no problem with articles about particular languages citing the ISO 639 identifier. It’s a way for people that may have a valid need to use an ISO 639 identifier to discover what it is. If there is a situation in which the ISO 639 identifier and the concept it denotes needs some qualification, that can be included in the Wikipedia article. What I don’t like is Wikipedia articles that include the code tables of ISO 639: besides being (at least in a moral if not legal sense) in breech of ISO copyright, it encourages users to refer to a non-authoritative source that may or may not be accurate.
So, in summary, I expect “most of us” will have situations in which it makes complete sense to use ISO 639, but certainly some of us will have situations that require something else.
One common misunderstanding is also found in this post: ISO 639-3 does not attempt to standardize languages, but language names. [jnj] thus does not refer to the language Yemsa, but to the language name “Yemsa”. I do not fully understand what that means (and why one would want to create codes for names in the first place), but at least this is what ISO/SIL purport to do.
To answer some ultimate questions that Morey postulates, the standardisation of languages by ISO is not only necessary, but a pre-requisite for creating, publishing and sharing electronic documents like Word, PDF, HTML or LaTex and e-mail files.
Without these codes, which are not used only for catalogues, the everyday electronic communication and production of documents would not work.
So yes it is useful, productive and more than appropriate for electronic communication -which is now the main form of distant communication.
I think Morey is not recognising this main dimension of usage of ISO language codes, when he is trying to compare it with standardising “religions or cultural practices”, which can’t be compared.
This seems confused. In how far is ISO 639-3 relevant for the creation of a text file? I would not even know which of the listed file formats support ISO 639-3. HTML for one requires RFC 1766, and email and LaTeX files are basically plain text without metadata where one could store language tags.
I am referring to the paper which relates to all ISO 639 language codes.
Most email nowadays is HTML encoded, and you have to do lots of encoding if you want that email-viewer e.g. displays the Subject in Traditional Chinese (used in Taiwan) correctly.
RFC 1766 relates to IANA which again leads to ISO.
@Sebastian:
HTML 4 was created many years ago, before ISO 639-3 existed. Web specifications have evolved since then. The HTML 5 candidate recommendation specifies for the lang attribute that a valid BCP 47 language tag must be used. The BCP 47 specification, in turn, normatively references ISO 639-3.
ISO 639-3 is, in fact, extremely widely referenced (directly or indirectly via derivative specifications like BCP 47) in information systems today.
@Stephen & @Sebastian N.
I am a little confused as to the the relevance of your reply remarks in the case of the ISO 639-3 code set. As apposed to other ISO standards the current ISO 639-3 tables are freely available from the registrars’ website. In contrast to the free and accessible nature of this available data (from the registrar), if one were to purchase the ISO 639-3 standard from the ISO, my understanding is that the ISO would only be able to provide the 2007 version of the standard which does not include changes the registrar has made since the induction of the the standard. ISO 639-3:2007.
Also to note: the registrar does make the requests for changes, and the changes public. (@Martin, which part of the process is not transparent enough for you? articulating the currently hidden parts would, I think greatly add to your argumentation, and bring the discussion in a practical and actionable direction.)
While arguments are being made for and against current practices with respect to registrars of ISO standards, and while I recognize that this post is mostly about the ISO 639-3, it is interesting to note that no remark has been made about the ISO 639-2 code set not being available from its registrar during the past U.S. Government shutdown (when the Library of Congress web pages were redirected to a “404 type” page). So this leads me to ask, are the arguments presented here purely made in hopes of improving the mechanics of the ISO system as they relate to standards with which linguists are likely to interact, or are they directly related to a particular registrar and a particular standard? If they are related to how linguists interact with the ISO in language based technical domains then it seems that the discussion should include elements/standards from ISO/TC 37/SC 4 where lexical data storage formats are outlined. To my knowledge these standards are not open (though if this is not the case, please point me to where I can access them). Nor is it clear how the discussion surrounding the formation of these formats/stnadards was conducted.
It is obviously nice that the registrar makes the standard (and updates) available. The fact that ISO does not include the updates, however, raises the question: what is the role of ISO in here in the first place? Apparently, they do not provide up-to-date standards; furthermore, they charge money for outdated standards. The whole process seems to work without ISO playing any role in there. If this is the case, one could just as well ignore ISO altogether (which is what I would recommend).
As for ISO/TC 37/SC 4, I would also argue for a free standard. The case here is a little different though, because ISO/TC 37/SC 4 has obvious applications in industry, whereas ISO 639-3 has none (besides trivial ones). ISO is not a good place to host standards for academia, but it might be a good place to host standards for industry (I have no idea).
Language identification is of course relevant in many commercial applications, but ISO 639-1 seems to do everything which is interesting in that domain.
Indeed, ISO 639-1 does not cover everything relevant.
Exactly 639-1 code is the example of the sentence from the paper: “ISO 639 has the potential to be misunderstood, misused and abused by governments and other decision-making bodies” – because one of criteria for 639-1 is that language must have recognised status in country, meaning ISO 639-1 is there only for languages recognised as such by governments, which can and obviously is used by governments to force one official language and actively neglect others, leading to endangered languages and to languages disappearing.
@Sebastian Nordhoff
With Respect to the ISO and the availability of updates (or updated code tables) from the ISO’s website (rather than the registrar’s website), I agree that there seems to be a disconnection, a gap, or a break in the distribution process. I am not familiar with how a registrar updates the ISO’s website so I can’t speculate where the breakdown in communications is (if there even is one). It may be that this is the responsibility of the TC 37 chair and has absolutely nothing to do with the ISO 639-3 registrar’s capabilities to distribute changes.
I agree that such a discrepancy (between the registrar’s web accessibility practice and ISO web accessibility practice) seems to communicate a less relevant role on the part of the ISO’s “authoritative” distribution platform. – Though I caution that one possible perspective here could be that the ISO views the registrar’s website as “official”. I am not saying this is the case, but I am saying that that it is a possibility. This communication gap and the faults thereof, should not be leveled upon the organizations who are the registrars of the various portions of the standard, but are reflective of the management practices of the ISO itself. – So merely changing the organization deputized to be the registrar of any section of the ISO 639 (but in this case the ISO 639-3 portion), is not necessarily going to solve any problem with respect to clear and effective communication surrounding the standards in the ISO – especially ones with which linguists interact.
With respect to ISO 639-3 relevance in industry, again, to push for clear and articulate argumentation, you say:
I find this statement to present a logical impossibility.
“trivial ones ≠ none”. That is, I don’t see how one can argue that the ISO 639-3 code set has no relevance in industry, and also has trivial relevance in industry at the same time. Either it has some relevance or it has none. If it has some relevance then, I think you are not appropriately framing the conversation.
Since you seem to concede that there is some (meaning more than none) relevance of the ISO 639-3 to industry, I think the issue you are trying to drive at is the applicability of the standard to different use cases within certain industry segments or within various diverse industries. You would have to confirm this, otherwise I am putting words in your mouth which I don’t really want to be doing.
You mention academia, when you say:
. There are several things which are not clear to me in this statement: (1) what do you mean by “academia”? – Do you mean all academia or do you mean a sub-section of academia, specifically linguists focused on typology, or field linguistics? (2) It seems that you are pre-supposing that “academia” – whatever is meant by this term – is the only users of this standard. To my knowledge, it seems that academia is representative of only one market segment (or industry segment) with an interest in using the ISO 639-3 standard. If I have my history correct the standard has never been “in the hands” of academia, but academics have always had input into the standard’s development and progression.
So, while you are correctly pointing to some of the shortfalls of the current system, I would hope that you might strengthen your argumentation by: (1) clarifying your terminology, use cases, and the relationships of the entities involved and how they relate to the standard, and (2) present some positive models for effective management, where the standard is widely adopted in various market segments and in various industries. In your above remarks you fall short of saying that the standard should be “managed” by academics or “the academy”, you also fall short of saying who should “manage” the standard. Your argumentation would be strengthened by providing some examples and models of success.
Note: You also do not define “free”. It would be helpful if you would also outline what you mean by “free”… As “free” can mean different things to different people, e.g. there is the BSD vs. MIT vs. CC-BY-SA-ND-NC versions of “Free”. When I say “free” in reference to the ISO 639-3 tables from the registrar I mean without cost to the consumer of the tables.
Disclaimer1: I do think that a list of language names and codes is useful
Disclaimer2: I have no particular problem with SIL’s background in this particular case
regarding 639-3 in industry: trivial use cases would be: this document is written in Dyirbal. More advanced use cases would be localizations of software packages and the like, machine translation and other NLP stuff. Let me thus rephrase my claim:”ISO 639-3 has no measurable impact on industry.” See http://aclweb.org/aclwiki/index.php?title=List_of_resources_by_language
Regarding your other questions
(1) academia was taken to mean: science, but not applied science. Anybody who does research without interest in exploiting the results commercially. In this context, I would count SIL as academia (and definitely not as industry).
(2) I would be interested in hearing more about other users of ISO 639-3 outside of academia.
(2bis)[positive models]. I would think that IETF would be such a positive model. In the more narrow domain of linguistics, the old Ethnologue codes, and the WALS codes would be in a similar vein. For the particular domain of typology, I prefer best practices rather than bureaucratic “standards”. In the long run, users will decide which practice suits their need best. The Leipzig glossing rules are also such a best practice. There is no way to enforce them, but they seem to be widely adopted.
ISO standards as I see them are there to assure that components from different manufacturers can interface in a workflow. If your printer wants A4 and the paper mill produces A4 without having to talk to the manufacturer of the printer, that’s obviously useful.
The output format of one tool can be the input format of another tool. By saying “My tool adheres to standard X”, people know how to integrate this tool into the toolchain. If they make an investment (new machines or the like), they can be sure that they can change suppliers etc if the need arises. I can see how this would translate to ISO/TC 37/SC 4. I fail to see how this tool chain idea would translate to ISO 639-3.
One use case of ISO 639-3 (or WALS codes, or glottolog codes for that matter) is pooling of resources. Say you have a phonological database and a syntactic database from different sources and you want to merge them. This is much easier if both share a vocabulary, e.g. the vocabulary for languages. I would advocate using Linked Data for this, and lexvo.org or glottolog.org would provide the vocabulary.
Our knowledge of language classification is likely to improve in the future. I see why one would want to have something like ISO 639-6 (and I also see that it is a disaster), or like Glottolog. Regarding update procedures, I would like to see something like RELcat for language classifications, i.e. a collaborative mechanism.
Regarding “free”: standards should be royalty-free (especially if the creation was funded by taxpayer money), and free to distribute. In the CC world, this would look like CC-BY-ND, although I think that CC-0 would be more appropriate as far as licensing is concerned. The BY part would be added anyway, and if you want to claim that you use the standard as defined BY, ND automatically follows.
@Hugh:
“With Respect to the ISO and the availability of updates (or updated code tables) from the ISO’s website (rather than the registrar’s website), I agree that there seems to be a disconnection, a gap, or a break in the distribution process. I am not familiar with how a registrar updates the ISO’s website so I can’t speculate where the breakdown in communications is (if there even is one). ”
There is no breakdown in communications: the ISO Central Secretariat does not publish current code tables for any part of ISO 639, and never has. The parts of ISO 639 that include code tables each establishes a registration authority to maintain the code table for that part of ISO 639, and it is the responsibility of that RA to make up-to-date code tables available.
@Sebastian:
“ISO does not include the updates…”
These comments suggest a misunderstanding of how the ISO 639 standards and ISO’s role in relation to those standards actually work. If you were to purchase a copy of ISO 639-3 from ISO, it would not contain any code table with language identifiers — it has never included them. Rather, the standard does establish the existence of a standard code table and it establishes a particular registration authority (SIL) for the maintenance and publication of that code table. That does not mean that the ISO 639-3 standard provided by ISO is not up-to-date: until it is superseded or withdrawn, it will continue to be up-to-date in what it claims to specify.
The situation with 639-3 is slightly different from that with 639-2: That document was written with a code table included as an annex. Now, the 639-2 code table has undergone changes, and so one could rightly state that the code table that appears in that document is out of date. But that is different than stating that the standard is out of date. Like 639-3, 639-2 establishes a registration authority (Library of Congress) to maintain and publish the up-to-date code table.
“ISO is not a good place to host standards for academia.”
ISO standards may serve some useful purposes for those in academia, but not necessarily all. Anybody in academia that uses HTML, Wikipedia, or any number of other examples of modern technology will be (perhaps unknowingly) using ISO 639 (and, in fact, ISO 639-3), and probably find those things useful. At the same time, it’s certainly true that academics (linguists, in particular) will have needs in some situations that cannot be met by ISO 639. That does mean that ISO 639 (including part 3) is not useful or not a good thing for ISO to be providing.
Thank you Sebastian for pointing this out. You are absolutely right, a good deal of money has to be paid to get the ‘standard’ from the ISO. This is indeed contrary to good scientific principles.
I am glad that this issue is getting some discussion.
In reality, this is a non-issue: While one might have to pay for the actual text of ISO 639 standards, it is not the text in the standard that is actually of interest for any real application. Rather, it is the code tables, and these have been made publicly available for use without fees in any application (except simple republication of the code tables).
Sebastian commented:
“Regarding ‘free’: standards should be royalty-free (especially if the creation was funded by taxpayer money), and free to distribute.”
I would point out that development of ISO 639-3 was not funded in any significant way by taxpayer money.
One major point why ISO is not a good place for scientific endeavours has not been mentioned: ISO standards normally have to be bought. The basic idea is that a group of companies creates a standard, and in order to be part of that group, you have to buy the standard. ISO 639-3 for instance can be bought for the price of 74 Swiss Francs. http://www.iso.org/iso/home/store/catalogue_tc/catalogue_detail.htm?csnumber=39534
I would very much advocate creating standards outside of the ISO world, since setting up more barriers to the spread of scientific knowledge (in this case monetary barriers) is certainly not what this world needs.