please revert to allow example namespace

[jonassmedegaard]

e9fe1e5 causes failures in several modules - quite likely the very same that Kjetil had special concern about: https://bugs.debian.org/922878

I will now revert that change for Debian, and suggest you consider doing the same upstream.

[kjetilk]

I suggest holding on for a bit, I remember having some very confusing issues around this, but I don't remember what it was.

[kjetilk]

Ah, found it, it was #21.

[kjetilk]

What happened here was that people would vote up certain prefixes on prefix.cc, and if you have e.g. http://example.org/ in tests that use URI::NamespaceMap's guessing code (which is a nice feature, saves you a lot of typing), tests will break in unpredictable ways due to crowdsourcing beyond your control, and depending on what modules are installed on the testing system.

I still don't see a good reason why you'd want crowdsourced prefixes for URIs that include the reserved domains, the low number of votes is likely to make this much more unstable than the more popular namespaces.

I can see several ways to fix this, one is to not have integration tests in upstream modules use the guessing code of URI::NamespaceMap. That would likely lower the total coverage. One is to invent totally random URIs for the tests that now use http://example.org (and friends). That would make the tests less readable, I think. One is to always enter those explicitly to the URI::NamespaceMap constructor, but that would make for more typing.

[jonassmedegaard]

a) Your patch is not limited to IANA reserved domains, but suppresses any domain containing "example".

b) Relying on Internet resources essentially is crowdsourcing ;-)

c) If you want non-surprising lists, then either fixate on a specific date, or use RDF::NS::Curated.

d) I still miss information about what actually breaks when using domains containing "example" compared to domains containing e.g. "foobar". I raised that question in #21 as well.

[kjetilk]

Long story short (since I spent too much time already): I don't think this should be reverted, as it solved a real problem. It is not ideal, so, I think rather than revert, improve.

[nichtich]

Would allowing example.org domains but including them from the crowdsourced list generated by https://github.com/nichtich/RDF-NS/blob/master/update.pl solve the issue? It's ok to define a prefix for example domains but this module should ignore what prefix.cc defines for them. Am 21. Februar 2019 18:49:00 MEZ schrieb Jonas Smedegaard <notifications@github.com>:

…

e9fe1e5 causes failures in several modules - quite likely the very same that Kjetil had special concern about: https://bugs.debian.org/922878 I will now revert that change for Debian, and suggest you consider doing the same upstream. -- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: #24

-- Jakob Voß via Android

[kjetilk]

Yes, that would work for me.

[nichtich]

I've released 20190227 with the exclusion of example.* TLD removed but their current prefixes have been locked from future updates as set in version 20170111.