Hyrum Wright (hwright) reported the following on the dev@ list:
I noticed when browsing a repository using the new ?r=REV syntax for
GET, that the links generated by the page link to HEAD, not to the
revision given by r=REV.
He proposed a somewhat naive patch, and discussion ensued. Here's the relavent
bits of a revised proposal:
Martin Furter wrote:
>
> On Thu, 14 May 2009, C. Michael Pilato wrote:
>
>> Ben Collins-Sussman wrote:
>>> On Thu, May 14, 2009 at 3:18 PM, Branko Cibej <brane@xbc.nu> wrote:
>>>
>>>> IMHO he should use the exact same query string as was on the original
>>>> URL. Anything else is guesswork, no?
>>>
>>> Er, this is tricky. Suppose I type in "http://host/some/dir?p=x&r=y".
>>>
>>> This wil locate dir@p, then follow the directory back in time to
>>> revision y (where it might exist at a different location).
>>>
>>> The children displayed in the output are therefore all being displayed
>>> at revision Y, and at some (possibly new) path. My first instinct is
>>> to point out that they should all therefore have the ?p=y bit attached
>>> to their hrefs, but this doesn't take into account the possibly new
>>> parental path. :-/
>>
>> You're right, this is tricky. But I think the solution is as simple as
>> behaving *literally* in the way we interpret the inputs semantically:
>> respond to PATH?r=REV[&p=PEGREV] with a "301 Redirect" whose
>> destination is
>> the resolved, real location (path + rev) -- that is, REALPATH?p=PEGREV.
>
> Shouldn't this be a "302 Moved Temporarily" since if r=HEAD it can point
> to a different path later if it is replaced by something else?
>
> (I guess if REV is not HEAD a 301 is perfectly fine.)
>
> Martin
Bug
Major