[OpenAFS-devel] Re: Gerrit upgrade complete

Jeffrey Hutzelman jhutz@cmu.edu
Sun, 22 Nov 2009 19:54:23 -0500


--On Monday, October 26, 2009 05:44:42 PM -0500 Andrew Deason 
<adeason@sinenomine.net> wrote:

>> In this case, though, I think Shawn pretty much has a plan to kill
>> off numerical change IDs.
>
> Bah.
>
>> The main reason for this is that when gerrit moves to git storage, the
>> plan is that it will operate as a  distributed system. Change IDs
>> which are particular to a single server won't allow that kind of
>> distribution to occur.
>
> Yes, the centralization of gerrit is what I thought would let them
> exist. If it's moving towards something else, then okay. I may find it
> annoying, but okay.


It's not just annoying; it's a serious usability problem.  There is value 
in identifying changes, patches, bugs, commits, etc with identifiers like 
small integers or short names, because those are easy to remember, 
recognize, and use in conversation.  A SHA-1 hash has none of these 
properties, and while it may work fine and even have advantages for 
following links in a web browser or using a git client, it is not 
appropriate for human communication.

I think in this case, usability is more important than having the code 
review system be distributed rather than centralized.

-- Jeff