[vworld-tech] Modern MUD server design

Mike Shaver shaver at off.net
Thu Jan 22 15:00:07 PST 2004

On Jan 22, Brian Hook wrote:
> I seem to be in the minority for liking JS, but unless its performance 
> has increased significantly it may not be feasible for the level of 
> scalability I'd like to pursue (1000+ users -- no, I doubt I'll ever 
> get 1/10th that, but I'd hate to find myself locked out of scalability 
> because of the choice of language made early on).

Increased versus what, I find myself compelled to ask?  Did you have
performance problems in a previous experience with the Mozilla engine?

I've worked with JS in a lot of different embeddings, including
high-performance e-commerce nonsense, and a well-architected embedding
can often perform very well.  (You want to initialize the
library-objects once and share them read-only or COW with your different
sandboxes, etc.)

Indeed, Brendan and the other Spidermonkey developers (it's a stretch to
include myself in there, but I try to find time) try to be very
responsive to performance problems (and API limitations) reported by
embedders.  That's where the "lightweight" function call infrastructure
and dependent strings came from, and in fact the original property-cache
system was developed to bring performance up for the e-commerce
embedding mentioned above.

("Benchmark-only" optimizations are a harder sell, since they tend to
not represent real-world applications, and solving for them takes time
and effort away from more important improvements, and can often hurt
those real-world apps due to code-size and runtime bloat.)


More information about the vworld-tech mailing list