- From: Moore, Jonathan <Jonathan_Moore@Comcast.com>
- Date: Sun, 3 Oct 2010 07:14:08 -0400
- To: Willy Tarreau <w@1wt.eu>
- CC: HTTP Working Group <ietf-http-wg@w3.org>
Perhaps a Link header with a new link relation like "newer"/"older" (to allow iterating both directions through releases) or "latest"? Jon ........ Jon Moore On Oct 3, 2010, at 2:04 AM, "Willy Tarreau" <w@1wt.eu> wrote: > Hi, > > it recently happened to me again that some people downloaded an old > deprecated version of some software I authored, installed it in > production and blogged about it with links to the deprecated software. > The link is basically a link to the tgz, such as : > > http://haproxy.1wt.eu/download/1.4/src/haproxy-1.4.8.tar.gz > > When the link points to an old version which contains at least 50 > known bugs, then I'm sure that in the following days the support > mailing list will receive several mails reporting known issues. In > fact, knowing that users will experience issues sooner or later > worries me much more than to think they'll send mails about that. > > I've already thought about replacing the file with a symlink to the > new version or to perform a redirect to the new version, but that's > really not friendly to admins or packagers who really need version > X or Y because they maintain it and apply the fixes on top of them. > And in general I don't like removing contents from the net, whether > it's buggy or not. > > Thus I realized that we might be missing a possibly useful feature. > What is needed to cover the issue above would be either to have the > ability to report an "obsolete" or "deprecated" status to the user > agent for the file being downloaded, or to report a Location header > indicating where to get the newer version, or maybe both. > > In both cases, the user agent would have two possibilities : > - either it is automated (wget, curl, etc...) and wouldn't care > about those headers. It could eventually log them if it provides > a log of the downloads. > > - or it would be interactive and controlled by a user (browser), so > it could report the status to the user and ask him if he still > wants to download that content, or prefers to be redirected to the > updated location (which may either be a page or downloadable > contents). > > Since the behaviour only depends on whether we're responding to a > machine > or to a human being, I don't think it can easily be addressed on the > server > side alone. > > So, for instance, we could have : > > HTTP/1.1 200 OK > Content-status: deprecated > Location: /download/1.4/src/haproxy-1.4.9.tar.gz > ... + requested contents > > Or perhaps : > > HTTP/1.1 303 Found > Content-status: deprecated > Location: /download/1.4/src/haproxy-1.4.9.tar.gz > ... + requested contents > > Or maybe : > > HTTP/1.1 200 OK > Content-status: deprecated; preferred=/download/1.4/src/ > haproxy-1.4.9.tar.gz > ... + requested contents > > I think the former would be better since it still allows server-side > intermediaries to rewrite paths as they do for many apps today. > > Have I missed something ? Would it be useful to work on something > like this, > and is it appropriate for the HTTP spec ? > > Thanks, > Willy > >
Received on Sunday, 3 October 2010 11:15:26 UTC