httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Marc Slemko <ma...@worldgate.com>
Subject Re: vetoing hide.h
Date Tue, 31 Mar 1998 06:36:53 GMT
Why do we want to make a huge change to the API, take weeks fighting over
exactly what should be what, make all previous modules completely
incompatible, and end up accomplishing almost nothing except make everyone
even more annoyed

ISAPI modules written for IIS v1 still work with IIS4.  Apache modules
don't work from week to week it seems.  Granted, those with the most
trouble are because they are more advanced things like mod_perl and
mod_php that need much more than many other APIs can give, but there are
bigger problems with the API than just how things are named.  This is not
the time.  No more architectural changes unless they are deadly necessary.

AFAIK, Apache does compile on platforms with relatively short symbol
uniqueness.

On Tue, 31 Mar 1998, Chuck Murcko wrote:

> Rasmus Lerdorf wrote:
> > 
> > If HIDE gets vetoed and it is accompanied by some massive changes to
> > natively change the names of all the API functions then I am all for it.
> > 
> That's the idea. It takes no coding discipline at all to write a
> function and name it 'write'. A now infamous so-called programmer once
> wrote an entire help package like that. Every one of his public symbols
> was a conflict. He showed no discipline, and what a crock it was. Palloc
> is a gimme; I'd wager the other two you've hit are also.
> 
> The realistic solution is to pick a unique prefix for public symbols.
> Two letters (the underscore counts, but only marginally) is not unique
> enough, especially if there's ever an Apache doodads library, or
> something. As inelegant as it may seem, cleaning up the prefixes is the
> correct solution. ap_srv_ probably isn't unique enough either (think:
> APplication SeRVer). You're allowed to be verbose; this isn't System 6
> anymore. Mix case. Do it once, and do it correctly. Commercial projects
> do this all the time. Add prefixes that clearly connote function. This
> probably should have gotten into the original coding standards, but, oh
> well.
> -- 
> chuck
> Chuck Murcko            The Topsail Group             West Chester PA
> USA
> chuck@topsail.org
> 


Mime
View raw message