loggerhead: alternative web front-end

Robey Pointer robey at lag.net
Sat Dec 16 00:09:52 GMT 2006


On 15 Dec 2006, at 10:19, Goffredo Baroncelli wrote:

> On Friday 15 December 2006 12:10, Robey Pointer wrote:
>> The code is hosted in a bazaar branch (of course) here:
>>
>> 	http://www.lag.net/~robey/code/loggerhead/
>>
>> And I'm hosting the bzr.dev branch through logger head here, if you'd
>> like to check it out:
>>
>> 	http://www.lag.net/bzr.dev/
>>
>> If anyone has suggestions, comments, flames, patches, etc, I'm open
>> for them.
>
> Really nice work; I like a lot the look and feel.
>
> However I have some suggestions:
>
> * In a web interface, the revno has low importance due to the its  
> inability to
> be stable between different branches.
> After some experiment, I reached the conclusion that it is better  
> to show both
> the information ( revid and revno ), because the revno represent  
> better the
> topology, and the revid is stable between the branches.

I knew that would spark some comments. :)

I had two issues with the revid, that ended up with me deciding to  
hide them completely:

1. A revid almost always contain someone's email address, which I  
really don't want to post on a website.

2. A revid is usually pretty long, and ends up looking like visual  
noise.

One tactic I tried at first was to SHA-1 the revid before displaying  
it.  While it was shorter and just as unique, it was also even more  
meaningless than before.  Two people I showed it to -- both strong  
coders but not using bazaar -- hated it and thought it was distracting.

I know the dotted-revno may change as the branch is updated, but I  
figured for a single browsing session, it will almost always stay the  
same.

If there's some way to show the revision id without actually  
displaying an email address, I'm open to it.


> * In history pages, in order to permit a better browsing I shows  
> also the
> following information:
>   - merge points
>   - the possible heads ( think about a branch in the future)
> (
> http://goffredo-baroncelli.homelinux.net/bazaar-dev/bzr-sha1? 
> cmd=changelog;otherrevid=;rev=john%40arbash- 
> meinel.com-20060817133350-22c93d5770f97d1b;pathrevid=;path= )

I saw this in the code, but didn't understand it.  Are those the  
merge points of the latest revision in the history?  (It made sense  
to me to show merge points and merged-from points in individual  
revisions, but I couldn't figure out what it meant for a history to  
have a merge point, unless it's just the latest revision's merge point.)


> * In the diff, it is no so obvious that the green lines are the  
> added lines
> and the red lines are the removed lines; is it possible to add the '+'
> and '-' simbol at the begin of the lines ?

What if I added a key at the top, like trac uses?  (It lists which  
color means what.)


> * When you display information on a file, you pass in the url the  
> revid and
> the path, so how you handle a rename case ( the same file may have  
> different
> name in differ revision )? In webserve, I pass the tuple:
> - path		-> the path
> - pathrevid     -> the revision related to the path
> - revid         -> the revision when you want to watch
>
> From the first two I compute the fileid.

Suddenly I understand this.  Thanks. :)  I guess what I should be  
doing is passing the file_id directly, and not the path.


> Regarding the copyright owner, I have to highlight that my work is  
> a derived
> work of the mercurial web interface, so I have also Matt Mackal as  
> copyright
> owner.

I'll add Matt Mackal (and apparently also Jake Edge, according to  
google) to the copyright notices, then.

robey





More information about the bazaar mailing list