Re: iovecs and grumpiness

[Greg Stein <gs...@lyra.org>]

On Mon, Jun 26, 2000 at 07:53:44PM -0400, Jim Jagielski wrote:
> rbb@covalent.net wrote:
> > 
> > The problem is that you and I are attacking this very differently, and I
> > have been unable to explain what is in my head.
> > 
> > I have reviewed your code, and I am trying to explain where it falls
> > down.  Obviously, I am not doing a good job.
> > 
> 
> Sometimes, the actual code makes it much more clearer... How about
> this: take the code-as-is, and then, if it's not too much work,
> create a small patch file that implements what you mean. Doesn't
> have to be the whole implementation, just enough to give everyone
> a taste. Of course, that may be a LOT of effort, in which case
> it won't work... Howz that sound?

Ryan is working on his patch. It should be available tonite.

I'm simply grumpy because I have been explaining the same thing over and
over again. It is like I'm fighting a set of blinders. To help out, I've
posted some working code. But even when I try to use the code to speak for
me (in case I'm doing a poor job of explaining), I am still banging up
against these blinders. The code is there, with examples to demonstrate, but
there is this new notion (from who knows where) that iovecs are needed. Or
that every layer must use a BUFF and do memcpy(). Of course, I'll take that
over the whole "recursion is bad because it causes register spill" FUD. At
least for the former, I can point to example code.

Now this could simply be marked up to something "deep" that Ryan hasn't been
able to explain. But I'm not seeing it yet, even with the filter examples in
front of me, and a rewritten mod_include in my head. Until those are
clarified, I'm just frustrated because the comments appear to have no basis.

I'd love to discuss the benefits/costs of my patch. But I'm feeling there is
more defensiveness on my part against incorrect characterizations and
assumptions, rather than an actual discussion or explanation. Maybe I don't
need to defend/explain, and just let the code talk.

Cheers,
-g

-- 
Greg Stein, http://www.lyra.org/