On Mar 8, 2004, at 2:04 AM, Carl Gutwin wrote:
> Karl Fogel wrote:
>
>> I'd be willing to answer some questions.
>
> Thanks very much! Based partly on what Ben Reser has already
> discussed, here we go:
>
> - Do you ever work closely on a feature or fix with another person,
> such that the mailing list or the issue tracker doesn't provide the
> right level of communication support? If so, what tools do you use in
> these situations to coordinate your actions? IRC was mentioned
> earlier, and I also see some indications that people get together face
> to face or on the phone. How about 'svn edit' style commands?
There are currently no 'svn edit' style commands, so that doesn't
figure in to our workflow ;-)
As for other types of communication, IRC, phone, and face to face
discussions do play into it from time to time, depending on the
situation. If several developers who are interested in an issue happen
to be on IRC at the same time they will discuss the problem there, and
if a problem is particularly contentious someone might try to work
things out via a phone call. Some developers work in close physical
proximity to each other, and others occasionally see each other at tech
conferences, so in those cases face to face meetings come into play.
That said, if I had to rank the different means of communication I'd
say email is the most used, followed up by IRC and the issue tracker,
and everything else comming in after that, at least for the developers
who are not in close physical proximity.
> - There is clearly a culture of making discussions public - does this
> extend to making sure that stuff from IRC (or phone conversations, or
> whatever) gets copied to the mailing list or the issue tracker? Is
> this issue of public/private conversations something that people have
> different opinions on?
If something is 'decided' over IRC or on the phone or whatever, the
discussion will generally be summarized and sent out to the mailing
list. This is necessary both to keep everyone else in the loop and
because those who were not present for the discussion still have a say
in what's going on, and the 'decision' isn't really final until we have
achieved some kind of consensus. Usually everyone else will agree with
whatever was decided, but not always.
> - I'm interested in the idea that developers don't really have a 'home
> base' and may work on anything that interests them at the moment - is
> the code particularly loosely coupled such that this can work? Surely
> there are efforts that have dependencies on other work that's also in
> progress. Is there usually enough up-front discussion that these
> dependencies get noticed before they become a problem?
Well, there are certainly some cases where certain developers are more
experienced in certain areas than others (Mike Pilato knows more about
libsvn_fs than most of the rest of us put together, for example, and
Greg Hudson wrote the ra_svn repository access layer), so sometimes
there is a clear, but unofficial, maintainer. In those cases it just
means there's someone who should probably take a look at whatever
changes you are planning to make. This isn't a hard requirement, it's
just good engineering practice, it's always better to get changes
reviewed by someone more experienced.
In other cases we do benefit from the fact that separate parts of the
codebase are largely decoupled from each other. There are clear
separations between the various libraries, and depending on what you
are doing you may not need to actually make changes to more than one of
them to add a particular feature or fix a particular bug. Of course in
some cases work can have a larger effect, touching multiple parts of
the system, and when that happens we deal with it.
In any case, large changes, new features, and stuff like that will
almost always be discussed on the mailing lists first, and depending on
the scope of the change in question the discussion could be quite large
and detailed.
> - When issues move from the mailing list to the issue tracking system,
> does all further discussion also move over as well? Is part of the
> reason for the move to reduce traffic on the main dev list? Do people
> treat the issue tracker as basically the same type of forum as the
> mailing list (but focused on the issue)? Have you ever run into the
> problem of an issue moving to the tracker, but then the conversation
> about that issue was not as public as it should have been?
Of course that kind of thing can happen, but since most (I think
anyway...) of the developers watch the issue tracker I don't think it's
ever been a huge problem. There is always the danger of having
important points discussed in the mailing list or the issue tracker
when most of the rest of the discussion is in the other, but in that
case a follow up email or entry in the bug db pointing people to the
archives or the bug db will do wonders for clarifying things.
-garrett
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Mar 8 13:41:05 2004