- From: Dirk Pranke <dpranke@chromium.org>
- Date: Fri, 21 Mar 2014 08:32:12 -0700
- To: James Graham <james@hoppipolla.co.uk>
- Cc: public-test-infra <public-test-infra@w3.org>
- Message-ID: <CAEoffTCFrHO8Wwm=mBVDCyOd_MMLS9spxQigOSyE3wMHGRozUg@mail.gmail.com>
On Fri, Mar 21, 2014 at 4:11 AM, James Graham <james@hoppipolla.co.uk>wrote: > On 21/03/14 11:03, Robin Berjon wrote: > > • The implementation report. That might be a bit trickier because I >> understand why you'd want to reference files, just as I understand why a >> year later someone might need to move those files around. I'm not sure >> how to handle this. The simplest option if for IRs to have a warning at >> the top stating that they were built from WPT commit deadb33f and links >> might go stale — whoever wishes to reproduce the same situation can do >> so by checking out the repo and reverting to that commit. There are >> other options but they're pretty heavy-handed so I'd rather not mention >> them yet. >> >> > Yes, I think implementation reports should specify which version of the > testsuite they used. I don't think that linking to the most recent version > of the tests on w3c-test.org is appropriate; if you want to precisely > recreate an implementation report you should check out the files and do it > locally. I consider w3c-test to be a convenient service to remove overhead > of casually running tests, not the source of all truth, so I don't think we > should complicate it to make unusual use cases easier. > If this is the purpose of w3c-test.org, then we should not allow specs or implementation reports to reference lints to tests or suites on it directly. Given that there already are specs that reference it, it one possibility would be to move w3c-test.org to a new name and provide a forwarding site on the old name that makes the old links work. I suspect that referencing GitHub repos directly might actually make the problem worse, as you'd then be assuming that (a) GitHub will be around forever and (b) you won't find some other service you'd rather use 5 years from now. -- Dirk
Received on Friday, 21 March 2014 15:39:41 UTC