httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alexei Kosut <>
Subject Re: unsigned host port
Date Mon, 07 Apr 1997 06:03:58 GMT
On Sun, 6 Apr 1997, Marc Slemko wrote:

> > I don't remember whether that bits being on indicates positive or
> > negative. If the latter, we probably get away with doing nothing. If
> > the former, we should definitely do something.
> High bit is set for a negative value.  For all values between 1 and 32768,
> the lower 15 bits in a 16 bit representation should be the same.  

Ah, okay. Thats what I'd hoped, I just couldn't remember.

> If an old module expected some particular behavior on a negative port it
> will perhaps work differently, but that is certainly outside the bounds of
> what the API defines.

True, but we should still consider bumping the API number, becuase it
will allow the developer to use signed/unsigned depending on the
version, and avoid compiler errors/warnings.

Dean: AFAIK, we don't increase the API number unless the API
changes... This makes sense to me, because there's no point
invalidating all the binary module files out there for no
reason. (Apache will refuse to load a module with a
MODULE_MAGIC_NUMBER that does not match Apache's current
value. Really, we need a CURRENT_API_NUMBER, and a
MINIMUM_COMPATIBLE_API_NUMBER, of which the former gets increased
whenever we make a change, and the latter whenever we make an
incompatible change. We'd check the latter before loading a
module. But I've been waiting on 2.0 for this, because without a good,
working dld setup, it's pretty much pointless anyway, because most
people will recompile their modules with every new release anyway).

Alexei Kosut <>      The Apache HTTP Server

View raw message