https://bugs.koha-community.org/bugzilla3/show_bug.cgi?id=38425

--- Comment #17 from Alexander Wagner <[email protected]> ---
Just to add to the discussion.

I admit that what I implemented are very well known ids in a _scientific_
context. (Probably openlibrary being a corner case in this regard.) Let's take
a look at DOI as the most well known kind of ID in 024 as a guinea pig here.
The others really work the same way.

> 1. The structure of the source code is explained by Library of Congress. The 
> documentation does not discuss a link.

Agreee, wrt the documentation at LoC.

But just printing these ids without a link behind them seems a bit strange for
an _online_ catalogue and in 2026 where we do have Internet almost everywhere
available all the times. Especially, if we look right now at a Koha landing
page. So, I do not see a point in asking the user to copy & paste said IDs and
transfer them to the proper resolvers manually. I mean, their purpose is to be
resolve not to be printed. In this regard they differ from ISBN or ISSN which
do not naturally resolve unambiguously.

> 2. The 024 tag does not include a way to customize the link with a $u (link) 
> or $y (link text).

Well, there is no need to customize anything here in the first place.

These IDs (at least the ones I implemented in the code) have exactly one proper
way to resolve and their value (probably with a prefix like `doi:`) is enough
for users in question to know what the links will do and where they will lead
them.

IOW cataloguing $u would indeed be wasting time and resources: it should
definitely be the same for all ids of a kind and adding some "flavour text"
doen't help the user either, but would rather confuse the issue. Again, if one
would really like this flavour text it would only make sense to add it
automatically so it is the same for each ID of a kind in every record.

Instead of printing the doi one could indeed create a link with a text like
`See resource at the publisher`. However, this would deny the user to use the
DOI e.g. to copy it to her literature management system or any other tool (e.g.
the local publications database or CRIS system) she wants to feed the DOI into.
So IMHO there is a reason to also print those ids literally and not hide them.

> That means there will need to be hardcoded links, which are difficult to 
> update and customize.

I agree. That's why I would prefer a hash that gives the resolver per id source
and that's evaluated in the code. It scales better and easier to maintain.

However, as mentioned, these resolvers are very stable even for a library. (For
IT, eternity is usually 5 years, so they live way longer than "forever".) These
resolver URLs are are _not_ URLs who usually are dead the moment you call them.
As mentioned, we do have these links in join2 for nearly 15 years now and the
updates I applied in join2 during that time were http://dx.doi.org/%s ->
https://doi.org/%s (the old link still works, we just adopted CrossRefs
recommendation for new systems) and to replace http -> https everywhere. In the
latter case also all http-calls forward silently to https. IOW if I would have
done nothing everything would still work. So, it's definitely not "adding URLs"
which would indeed be a maintenance nightmare. So one could also live with some
hard coded urls.

Long story short: the code could be improved by said hash, but its not a deal
breaker in my opinion.

> Would the link would go to the primary domain of the source, and not directly 
> to the item?

Usually, they link to the full texts (doi, arxiv, pmc) or very well known and
established free literature databases in the field that offer additional
services beyond the scope of Koha (eg. pmid for all medicine related sciences).
So they do add quite some value to the user in areas where they are applicable.

One could consider to extend it to persistent ids minted by open access
repositories, e.g. handle used by DSpace repos. I also mentioned urn in the
original issue, however to my knowledge there is no _universal_ resolver for
URNs so they failed to fit the bill.

openlibrary is probably the "worst" of the bunch, where I also don't know the
stability. (AFAIK it is pretty stable though.) I added it as there was an
explicit request to add openlibrary ids also to fetch covers. In my setting
this is a bit out of the scope, but in my catalouge we just don't have
openlibrary ids so issue solved. :)

BTW: if we could somehow introduce the mentioned hash that tells the code
`resolve x -> by y` one could pre-seed it with all well known ids and
individually "switch off" unwanted ids easily by removing the resolver from the
config. That's how I did it in join2, but all our instances provide all
resolvers so one might also call this over-engineering.

> 3. If the code functions as the link text, the code may not convey enough 
> meaning to the average user of what it is.

I would disagree here.

Sticking to the doi, I do understand that this would not convey enough meaning
in e.g. a childrens library, of course. But you'd hardly ever catalogue a doi
in this setting anyway. It will pop up quite a bit in scientific libraries and
your scientists will know the doi (and(!) wonder quite a bit if you just print
them and not add the proper link to the full text underneath).

> 4. There are multiple fields, such as topical terms, that have a $2 source 
> code. These do not display or function with links.

Which may be a pity for an online catalogue, but out of the scope for this
issue. ;)

Still note the slightly different meaning of source here. In 024 we can use the
source in quite a few cases to construct an unambiguous link. So they convey a
way to construct links to a resource, I do not have locally available. One
might consider adding those links along the links generated via 856, if this is
visually preferred. 

Still note that for a proper complete citation the doi should also show up with
the citation. So these days it should read something like (style details may
vary):

    Physical review letters 136(4), 043001 (2026) [10.1103/bvqj-pm3n]

and not only

    Physical review letters 136(4), 043001 (2026)

and leave the user to resolve the bibliographic blurb even though there is an
unambiguous link via the doi.

For topical terms OTOH, eg. if we know they are from Mesh this would not work.
They do not point to "the resource(tm)". One could construct a link to Pubmed
based on Mesh-terms that triggers a search there, which would be quite helpful
for some, but also compete with the link of the topical term within your own
catalogue. Hm. One could use LCSH to link to the LoC catalogue. Probably
interesting, but I'd have to cross the Atlantic to make use of the resource
just discovered.

-- 
You are receiving this mail because:
You are watching all bug changes.
_______________________________________________
Koha-bugs mailing list
[email protected]
https://lists.koha-community.org/cgi-bin/mailman/listinfo/koha-bugs
website : http://www.koha-community.org/
git : http://git.koha-community.org/
bugs : http://bugs.koha-community.org/

Reply via email to