[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Using urlcheck - how many broken links should we accept?

On Thu, Mar 11, 2021 at 10:57 AM Andrew M.A. Cater wrote:

> There are lots of websites that return error codes. There are still lots
> where an http -> https substitution would solve a missing website.

The https-everywhere rule-set might help to solve the https problem.


> There are lots of apparent errors where, in fact, it's just a directory level
> move - https://www-master.debian.org/build-logs/urlcheck/MailingLists

That is weird, /intro/cn is a top-level path, not a sub-path of
MailingLists. Probably there is a weird symlink on the server

> How much breakage should we accept for old sites / sites that have been
> renamed or simply no longer exist?

Depends on how much you can be bothered to fix, since the web changes
often there will always be newly broken URLs.

> For, for example, DPL nominations on debian-vote which list resume information
> / university "stuff" - is it worth going to fix old links?

I think historical pages (security updates, News, vote pages etc)
should either remain broken or get their links pointing to
archive.org, possibly automatically at the start of a new year.

> I'm very wary of global search and replace through all of the webwml. How
> much checking should we devote to seeing whether websites exist or still
> resolve?

Global search/replace for http -> https for domains with verified
stable working TLS seems safe to me, and a good idea. Same goes for
things that moved from debian.net to debian.org or similar. Same goes
for things that changed their URL layout slightly. I've done that sort
of thing in the past, using smart_change.pl.

With the new translation-check headers this becomes a lot harder to
do, since if someone pushes a commit while you are reviewing your
changes, you can't just rebase and continue your review, you have to
delete the translations update commit, re-do it and restart your



Reply to: