loggerhead navigation

[Michael Hudson]

James Westby wrote:
> On Wed, 2008-03-12 at 10:48 +1300, Michael Hudson wrote:
>> Slightly, I think.  I probably was a bit unclear about what was stumping 
>> me.  Let me try again :)
> 
> Thanks for the clarification. I think we were both a bit vague.
> 
>> I think it's pretty clear that when you click on a revision number of a 
>> change, you should go to the revision page for that revision.
> 
> Yes.
> 
>> Where I get confused is that I was I thought I could make navigation in 
>> loggerhead clearer by always placing revision views in the context of a 
>> line of revisions -- and I don't know what that would be for this view.
>>
> 
> That sounds like it could be a useful idea.
> 
>> I'm not sure one can say "line of revisions could be the normal list of 
>> revisions" because the normal list of revisions is the mainline for the 
>> branch, and the revision you click on may well not be a mainline revision.
>>
>> You could expand the definition of "normal list of revisions" to be "all 
>> revisions in the ancestry, merge sorted" but I'm not sure that would 
>> make sense for the other views.
>>
> 
> That is what I had in mind.
> 
> I tend to think in these terms, but I understand that others may
> prefer thinking mainly in terms of mainline revisions.

Right.  The approach I've taken in my prototype isn't totally restricted 
to mainline revisions, but considers a line of development to be 
something that you get by repeatedly taking left hand parents (so if you 
start with a non-mainline revision, you in effect walk back along its 
lefthand parents until you get to the mainline, then follow the mainline).

This approach seems to work well for PQM managed branches like bzr.dev, 
where each mainline commit is a defined feature but you can still 
examine the development of each feature.

> Can you expand on what you mean by "I'm not sure that would make sense
> for the other views." please? Is this because you are trying to
> make things like the main log mainline focused?

Yes.  For better or worse.

> I guess in a lot of views the line of revisions you are thinking about
> is defined by the view that you just came from.

Er, partly.  Generally it should be fairly obvious when the list of 
revisions changes -- clicking on a "view changes to this file" link, for 
example.  In other cases, you're just navigating around this list.

> However in this
> case we just came from a single revision of a single file. Do you
> see this as the main problem with picking a line of revisions?

No, not really.  What is more the problem is that I can't think of a 
line of revisions that fits the bill.  You _could_ just take the line of 
revisions that starts with the revision specified, but that seems 
arbitrary and not-useful (I think this is what loggerhead in effect does 
now, but it's a bit obscured).

 > I guess
> using the line that bought you to the annotate view is going to be
> confusing, as it reaches too far back in history.

More than that, the revision that changed the file isn't in general a 
member of that list.  The navigation could say "revision 2.5 of 123" I 
guess :)

> I'd like to hear the input of others.

Me too!

 > I heard some dissatisfaction
> expressed at the sprint about the interface of the current breed of
> web viewers.

Well, certainly I don't like any of them very much.

 > I didn't hear any specific complaints though,

That's the hard bit alright :)

 > and I missed
> any session that was held on the topic. I don't find most of them too
> bad, so I feel a little in the dark, and uncomfortable giving my
> opinions, as I feel I may be missing something big.

Well, unless other people chime in, I'd be most grateful to hear your 
opinions.  If I think they're inappropriate, I'll feel free to ignore 
them :)

Cheers,
mwh