This is an archived snapshot of W3C's public bugzilla bug tracker, decommissioned in April 2019. Please see the home page for more details.
LinkChecker should behave sensibly for news:// and irc:// URIs and not always show errors.
Link checker's URI scheme/protocol support comes from the libwww-perl library. At least in theory, all schemes supported by it should be also supported in the link checker. Now, regarding the cases from your example URL: news://news.mozilla.org/netscape.public.mozilla.svg This URI is not RFC1738 compliant. There's no hostname part in news: URIs according to the RFC. Maybe it should be nntp://...? news:netscape.public.mozilla.svg This looks ok, but the link checker @ qa-dev.w3.org does not have a default NNTP server configured. Olivier, is there one @w3.org or somewhere we (c|sh)ould use for qa-dev and v.w.o? As far as I know, there's no standard for irc:// URIs. That's probably why it isn't supported in libwww-perl (and consequently nor in link checker). I suggest you contact the libwww-perl authors if you feel it should be supported. http://www.iana.org/assignments/uri-schemes http://www.w3.org/Addressing/draft-mirashi-url-irc-01.txt (expired 8 years ago) Finally, news: and nntp: URIs are not too well supported in libwww-perl. In particular, they will only work sufficiently for link checker use for URIs that identify an article; the implementation for URIs identifying a newsgroup is pretty incomplete. Yes, that means that even if the news->nntp "bug" would be fixed and a default NNTP server configured for qa-dev and v.w.o, your URIs (both of which identify a newsgroup, not an article) wouldn't be completely supported by libwww-perl (nor link checker) at the moment. (This made me remember that I have some patches that I should send to libwww-perl upstream, those make the nntp support a bit better.)
I suggest making news: and nntp: explicitly unsupported.
For good, or for now (if/until LWP gets proper support for them)?
Well, v.w3.org should only report links as broken if there is no doubt that the link is broken, for most news: and nntp: links it cannot know that. We might support nntp://example.org which could check whether there is a DNS entry for example.org and maybe even whether example.org accepts connections to :nntp, but we should not do much more. As not supporting them at all is easier...
I just sent a patch to the LWP mailing list which makes the news:/nntp: support in it complete enough for link checker use, excluding the case where the NNTP server's hostname is not present in the URI. And that case and how to work around it has been documented in the link checker documentation for some time now. If my patch is accepted to LWP, I'm not in favour of dropping the feature altogether. We could possibly add a special message and short-circuit without even trying for the no-server-in-URI case. Or do that only if a default NNTP server is not defined, but the problem with this approach is that Net::NNTP's default server resolution is somewhat hairy and not exposed through the API. Keeping the feature requires no changes to checklink, removing it does :)
My news/nntp patch is included in LWP >= 5.804, so as far as I'm concerned, this bug can be resolved by upgrading LWP on v.w.o. Reassigning to Olivier for that. Note: I haven't personally tested libwww-perl >= 5.804.
Currently, all that's available on any debian distrib is libwww-perl 5.803-4. http://packages.debian.org/libwww-perl For the sake of maintenance of the validator servers, I'd like to wait and see whether/when 5.804+ gets into debian.