It's not just IRML. We've had several discussions about the callout
protocol, data encoding, and performance, for example.
It's also relevant to policy rules carried in content,
as I've proposed we do for data integrity in a multi-party environment.
Hilarie
Yang, Lily L wrote:
That's a good point. Our rule engine implementation did exactly that -- rule
base is already loaded into an internal format that is highly optimized for
rule matching.
Lily
-----Original Message-----
From: Ng Chan Wah [mailto:cwng(_at_)psl(_dot_)com(_dot_)sg]
Sent: Tuesday, March 26, 2002 5:46 AM
To: Mark Nottingham; Hilarie Orman, Purple Streak Development
Cc: ietf-openproxy(_at_)imc(_dot_)org
Subject: RE: About XML performance
IMHO, it does not matter if the parsing of XML is slow. Normal
implementation of IRML engine will not parse the rules in XML format
everytime a request/response passes by. Typically, I will expect
implementors to parse the IRML only at load time, and store
the parsed rule
in some native format for faster retreival.
XML parsers have come quite a way. Some of the newer implementations
(esp. MSXML4, IIRC) can achieve astonishing results. XML
Pull Parsers
are also very good, depending on your application.
Saying "XML processing is very slow" is somewhat like saying
"intermediaries are evil"; it depends. And I'd note that
his product -
an XML accelerator - could be the reason that he's saying XML is so
slow ;)
On Monday, March 25, 2002, at 03:46 PM, Hilarie Orman,
Purple Streak
Development wrote:
Though it is primarily a product announcement, this article has
some interesting performance claims about XML:
http://zdnet.com.com/2100-1107-866560.html
Short story: XML processing is very slow; it becomes the
bottleneck
for a website.
Hilarie
--
Mark Nottingham
http://www.mnot.net/