axis-java-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sam Ruby" <ru...@us.ibm.com>
Subject Re: The Great Debate: Xml Parsers
Date Thu, 22 Mar 2001 13:39:03 GMT
James M Snell wrote:
>
> As Glen Daniels pointed out in his earlier message, "getting the
> product out" is perhaps the most important thing we need, however,
> we need to try to do so in a way that gives us the best performance
> possible.

+1.  It is worth noting that Axis is essentially V3.  We were able to get
V1 and V2 out on top of Xerces.

> We need excellent performance on small documents

I've said it before, and I will say it again.  I believe that the impact of
performance of the XML parser for small documents is overstated.  As Ted
has pointed out, there already is a lot of overhead associated with doing
things like RPC over text files over http over sockets.  Optimizing the
last 3% out of an area which is not where the majority of the time is being
spent is not a good use of our time.

As evidence, I don't recall anybody remarking on any noticable improvement
when converting over to JDOM.

> and a very low memory and installation footprint (for
> resource-constrained devices).

I do not see this as valid considerations for the server.  There will be a
wide range of requirements for the client, and the axis code base need not,
and in fact can not, address them all.  Axis needs to focus on the 80%
case, and leave the rest to special purpose clients.  That's why
interoperability is so important.

> I agree on your long term perspective -- Axis is an Apache project and
> should be built using Apache tools.  Let's start putting our heads
> together to figure out what we need to do.

+1

- Sam Ruby


Mime
View raw message