Re: [HACKERS] Developer's Wiki

Tom Lane wrote:
> Martijn van Oosterhout <kleptog(at)svana(dot)org> writes:
>
>> On Sat, Sep 16, 2006 at 09:15:24PM -0400, Tom Lane wrote:
>>
>>> Fortunately, none of the real developers would have to pay any attention
>>> to any such page ... and you can bet they wouldn't.
>>>
>
>
>> If someone wants to spend an afternoon putting up a coherent
>> description of their wishlist item complete with possible problems and
>> solutions, then I don't see why we should stop them.
>>
>
> Because if they're willing to put any actual effort into it, the right
> way is to post that same item to the mailing list where it can be
> discussed. If it survives such discussion (very possibly in a modified
> form) *then* it belongs on a TODO list. The first problem with a wiki
> TODO is that it will not reflect any sort of community consensus, only
> the opinions of whoever edited the page last. The second problem is
> that setting it up represents a unilateral attempt to redefine (bypass?)
> the community's design/development process, which is a process that has
> served us well for many years and is not showing any signs of being
> broken.
>
>
>

I agree with lots of this.

Being slightly more abstract, we are grappling with a couple of
different kinds of objects here: discussions and decisions. The mailing
list is a very good way of having a discussion, and a wiki is IMNSHO a
poor substitute. Ditto, bulletin board, web forum, blog ..... The
reason is simply that with a mailing list all you need is a subscription
to get the info delivered to you in a medium everybody uses. It's push,
not pull, and that's very appealing. Any other mechanism requires the
user to seek the location of the discussion actively to some degree.
Conversely, the very unstructured nature of the mailing list(s) makes
them a poor medium for capturing decisions. That's why some of us have
advocated use of a tracker to capture decisions about development
directions, because the TODO list doesn't seem appropriate. But an open
wiki would be a horrible substitute for the TODO list - it would turn it
from a list that reflects at least some discussion and consensus into a
mere wish list of no authority whatsoever. IOW, it is the exact opposite
of the direction I believe we should be headed.

I use wikis in my work as a good way of capturing all sorts of
information I want to keep. But I have generally found them to be less
than successful as a way of capturing discussions or developing coherent
bodies of technical information and decisions. Comparisons have been
made with WikiPedia - they are inappropriate. Quite apart from anything
else Wikipedia survives through the work of a huge team of editors who
review the work of contributors. And they still run into trouble. We
don't have the resources and we don't need the fights. So let's not go
there.

The only good purpose I can see for a developer wiki is as a place to
publish information that is too large for the mailing lists. Currently
we provide web and other space for a few users - a wiki would allow us
to provide publishing facilities in a central spot for a significantly
wider group of people, with very little cost.

Tom proposed a modest roadmap type experiment a week or so ago. I'd like
to see that pursued. After all, we know of some things that are at least
at first cut stage for 8.3, and a few things high on may people's
agenda. I'd also like to see some work done on using a tracker (for
features as well as bugs). The rest of what's been talked about strikes
me as wasted effort, to be honest. We seem to be running in a few
directions which look like dead ends to me. Let's pick one or two
strategically, and follow those instead.

cheers

andrew