Thérèse Godefroy wrote:
> The old website, which went down in August 2015,
> now points to (,

That seems like a very confusing configuration.  Because then a user
would find themselves on what looks like Savannah but by a different
domain.  And because of the domain difference it will use a cookie for
the other domain instead of the Savannah one.

Perhaps we should put in a redirect in that case so that if someone
arrives at Savannah by an unexpected and undesired name that they get
redirected to the main domain.  That would clean up this current
situation.  I think that would make the best of the situation since we
can't prevent the world from pointing to it.

I will make that change sometime this week when I get enough time to
make the change and test it.

> What we would like to do now is redirect all requests for
> http(s)://* or http(s)://* to
> http(s)://

I think you mean this address:

As far as I can see it would be the first project to have such a
redirect set up on Savannah.  Therefore I am immediately suspicious of
the new feature.

I think it could become an unmanageable mess if many projects were all
needing to be configured with redirects from Savannah.  Therefore on
the basis of cofiguration mess I would rather not see this added.
Simple is best and this is a creeping feature.

But it does appear to be an official GNU project.  It is technically
possible.  So I won't block this from happening.  And many things
about the GNU project require that someone manually edit files.  So
this wouldn't be that much different in spirit.

But I would want to be convinced.  Please say a few words in favor of
why this feature needs to be created for the first time for this
project?  And if so and if it is a desirable enough feature then I
could see that we would want to make that possibility available to all
other GNU projects as well.

Note that it will require a test for this feature to be added to the
test project as well.  If a test is not added then there is no
guarentee that it will not be broken at some time in the future
because it isn't otherwise documented.  After a few years (months?)
everyone will have forgotten about the configuration and then it would
be lost and broken.  I am trying to think of a good way to have a list
of available redirects available for download such that the entire set
could be automatically tested.  Because trying to keep the test
framework updated would otherwise be problematic.  They would probably
always lag behind and be out of date.

It also looks like email address for the registered
user dyfet must be broken because the mail would attempt to connect to
Savannah which does not receive email.  That needs to be updated.


Reply via email to