From: David Abrahams (gclbb-jamboost_at_[hidden])
Date: 2003-09-26 09:29:43
Vladimir Prus <ghost_at_[hidden]> writes:
> To a certain degree you're right. Though from my standpoint the
> situation looks a bit different. Several persons send patches for
> bugs. This tells that newcomer can locate and fix a problem.
Yes, that's sometimes possible.
> I don't know why nobody did any big changes. That may be because of
> lack of time, or because people don't run into design issues. I
> don't know, but it would be great to find out.
I can only speak for myself, but from my standpoint it's mostly
because the architecture of the system and its overall flow of control
and data is either unfathomable or just too complex to change with any
confidence. Having tests helps, but doesn't help very much if you
think it's likely you'll just break them no matter what you do. I
liken it to the structure of GNUs, which amazingly (to me) has several
developers working on it.
My pet peeve, property-adjuster, is an example of something incurs a
great deal of complexity in the architecture for a very general
mechanism which is only used in one specific way and for a single test
case which it's not very clear should be supported. If similar
choices are made in enough places, it becomes very difficult to hold
the design of the system in one's head. What little understanding of
the architecture I am able to develop dissipates quickly. If you have
to crawl through lots of code just to put together a picture of what's
happening before you begin, it becomes a poor investment to try at
-- Dave Abrahams Boost Consulting www.boost-consulting.com
Boost-Build list run by bdawes at acm.org, david.abrahams at rcn.com, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk