HTTPS on docbook.org is breaking stuff

[Shaun McCance <shaunm@gnome.org>]

I maintain a collection of tools in GNOME that includes yelp-check,
which has a validator that can automatically detect the input type and
do the appropriate thing to validate. By default it uses xmllint, but
it has an option to use jing.

At some point docbook.org switched to using HTTPS, redirecting any HTTP
requests to HTTPS. Great move, in general, except it broke my validator
and I'm not sure how to fix it. Rundown:

* Neither xmllint nor jing seem to follow the redirect. They end up
retrieving a file that says:

    Redirecting to https://docbook.org/xml/5.0/rng/docbook.rng

You can get a file just like this your self by calling:

    curl -O http://docbook.org/xml/5.0/rng/docbook.rng

Passing -L to curl gets you the right thing. But yelp-check isn't using
curl. It just lets xmllint/jing do HTTP.

* If you have catalogs installed correctly, there's no problem with
xmllint. I don't even know how long this has been a problem, because I
have catalogs. But jing doesn't seem to pick up the catalogs, at least
as they're packaged on Fedora.

* If I were to change yelp-check to use the HTTPS URL, then jing works
fine, but xmllint doesn't. I think xmllint doesn't support HTTPS.

* Also, if I were to switch to using the HTTPS URL, then catalogs would
no longer work, because they arean't matching on that URL.


So, basically, to get yelp-check working correctly, I'll need to detect
one of three cases and do something different:

1. using jing: use the HTTPS URL
2. using xmllint with catalogs: use the HTTP URL
3. using xmllint without catalogs: I guess curl the schema and just
hope it doesn't use includes because those will break if I curl it.

Is there any way docbook.org could just serve the schemas over HTTP as
requested? These URIs are kind of API, and the redirects are kind of
breaking that. xmllint and jing are both popular tools. I can't be the
only person running into this.

Thanks,
Shaun