Re: Comments on HTTP draft [of 23 Nov 1994]

> > It's actually better to ignore the IMS header and have the cleint cancel 
> > if the date doesn't match. 
> 
> IT MOST CERTAINLY IS NOT BETTER TO DO THAT!!!!!!!!!!!!!

My words exactly since a boldface font is not available.  Servers are
already facing problems with TCP kernel bugs, and intentionally
dropping connections from the client side would only make things
worse.

> The point is to save network traffic, NOT make life slightly easier
> on server implementors.  Supporting IMS is TRIVIAL and has already been 
> done on all major servers -- not supporting it is reprehensible and
> deserving of public abuse.

Yes; even proxies do it (at least both that I've written), and it was
doable even then, although slightly more complex (with all the
combinations of incoming and outbound requests).  But the benefits far
outweigh the complexity.

Cheers,
--
Ari Luotonen				http://home.mcom.com/people/ari/
Netscape Communications Corp.
650 Castro Street, Suite 500
Mountain View, CA 94041, USA

Received on Friday, 2 December 1994 12:22:20 UTC