Re: [Monotone-devel] Re: comments on win32 patches

[Jon Bright]

Hi,

Nathaniel Smith wrote:
> This feels like one of those places where special-casing isn't worth
> it, because while the special case is simpler, you could just use the
> general case, have better test coverage, and simpler code overall.  I
> could be wrong though, if Unix's path semantics are very complicated
> (which seems unlikely?).

My plan was just to go with executing 'which' from the Unix side instead 
of from the LUA side, and let the path semantics be handled externally.

> On the contrary.  Deleting code means you don't have to maintain it,
> and reduces the interlocking constraints in your system, giving you
> more freedom in implementing other things.  Usually, these factors
> free up much more time than you (might, possibly, maybe) save by not
> having to rewrite the code later... especially since you don't
> actually have to rewrite it, seeing as we have this nice VCS all the
> code is in ;-).

My comment was more with the article than the principle - but I'm all 
for deleting lposix if we don't need it.  Less because of the 
maintainability, more because of the potential security/correctness 
implications which it introduces (like the chdir() thing we discussed 
the other day).

--
Jon Bright
Silicon Circus Ltd.
http://www.siliconcircus.com