[Lmbench-users] Re: LMbench as gcc performance regression test?

Larry McVoy lm@bitmover.com
Tue, 4 Nov 2003 09:56:03 -0800


On Tue, Nov 04, 2003 at 08:07:50AM -0800, Dan Kegel wrote:
> Robert G. Brown wrote:
> >P.S. -- the xml-wrapping of the results is pure religion on my part, but
> >I REALLY URGE its adoption.  For that matter, I'd make the configuration
> >file xmlish as well.  If XML were universally adopted for system tool
> >APIs with a bit of sanity in its application the world would become a
> >Better Place (tm).
> 
> I could never understand why people preferred XML to lisp-style expressions.
> We're probably going to use lispy syntax at Ixia for our data interchange.
> So instead of
> <foo><bar>1</bar><bletch>2</bletch><name>Joe</name></foo>
> one has
> (foo (bar 1) (bletch 2) (name "Joe"))
> It's logically equivalent, but far less fluffy for most data.

XML in general seems to be one of those over hyped, over used and solutions
in search of a problem technologies.

To the extent that I have any say over the development of LMbench I'd resist
the introduction of another dependency.  Flat ascii files have worked for 
years and I don't see adding XML as doing anything useful.

On the other hand, if someone else thinks it is wonderful and they want 
access to the raw data we can provide a tarball or some other way of getting
that and they can do whatever they want.  
-- 
---
Larry McVoy              lm at bitmover.com          http://www.bitmover.com/lm