[
Lists Home |
Date Index |
Thread Index
]
One answer, Derek: COTS.
We have to do it that way. As I replied to Dare, we
prefer these be based on standards and that probably
means FastInfoset because there aren't good alternatives
that will be available in the timeframe alloted or
possibly ever given positions taken publicly by MS, IBM,
the W3C, etc. So it's off to ISO.
XML will be replaced. Technological churn is ceaseless.
You and I are both too close to the current winner to
know what that will be, but we are also old enough to
have seen fish become mammals when the seas boiled. At
that point, it will be another 'good-enough' solution
quite probably, and it will likely be a solution that
we are rejecting today because we don't need it yet.
Actually, we do use CSV for many things. Some customers
prefer it. XML is there for those that don't. Where
we have to use binaries, we can do exactly as you suggest
and have. We want a standard for exactly the same reason
as we use XML: the customer demands it and we don't want
to lose business explaining YAGNI to them. In a business
deal, the real technology debate is the last debate and it
is had after the contract is signed.
My suggestion to you, Joe, is to benchmark web service
performance. Consultants inside 395 need to come to
grips with these issues. Good enough isn't always.
No slur intended, at all. The closer any system
comes to real time, the more it matters.
len
(also my own opinion and not that of my employer)
From: Derek Denny-Brown [mailto:derekdb@microsoft.com]
So create a format tuned for your scenario, and stick XML APIs on both
ends to support plugging into existing XML machinery. This begets the
whole binary-xml question. The issue is not one of making a faster
format; that is easy. The issue is making a format that is enough
faster in enough different scenarios, to make it better than
individually tuned formats. CSV may be optimal for you, so use it. As
has been brought up, there are cases where CSV doesn't work (embedded
commas, newlines, nested structure, etc). As Dare commented, XML isn't
about being the best at anything. It is about being the good-enough for
the widest variety.
> Still, what will replace XML? What will it look like?
> Is it hiding in the tall grass or is it laying on
> the shelf at Sun? It's a fun question.
I don't think there is anything out there today that will 'replace' XML.
To replace XML it would have to provide enough benefit in enough
scenarios to justify moving existing systems. For most users, none of
the current proposals for something XML-Like (be it simplified XML or
binary-xml), are really necessary. XML's reach is already being
extended by using custom alternate formats in tightly coupled systems.
That trend is likely to continue, but for the near term, it is likely to
exist as an optimization path. XML-text will still be the primary
model. That model isn't likely to change until enough XML is rolled
out, that XML developers start to really become fed up with XML's
limitations due to it's origins as a 'document' format. Why can't a
user escape arbitrary characters both in content, but also in names? As
XML becomes increasingly used to communicate between systems, which do
not share its origins in text markup, these will become increasingly
problematic issues.
Just like XML is a SGML, generalized for the WWW markup, someday we will
see a new text format supplant XML that is generalized for _data_
exchange... or whatever need is found that XML does not quite fit.
-derek
Disclaimer: the above is my personal view and is independent of my
employer
|
