[
Lists Home |
Date Index |
Thread Index
]
On Mar 18, 2004, at 5:03 PM, Joshua Allen wrote:
>
> Honestly, the talk about "*real* pub-sub network" reminds me of the
> people in the early days of the Web who argued for a "*real* hypertext
> network". There were lots of *real* hypertext systems before the web,
> and plenty since, but the web is vastly superior.
I think it's very important to keep this in mind during this type of
discussion -- it may be that Metcalfe's Law will trump all the
arguments against RSS+polling+Aggregators as a universal pub-sub
network. I wouldn't be astonished if that happened, given the success
of the Web and all the other technologies whose ability to hit the
80:20 point outweighed their technical limitations.
BUT it's important to remember that "real" pub-sub systems, AKA
message-oriented middleware, are pervasive in enterprise computing
today. The enterprise infrastructure accommodated the Web fairly
gracefully because in-bound requests could be directed via CGI, etc. to
legacy systems to generate responses. After all, everyone understands
request-response. I'm not sure sure that the Web will prove to be a
good fit for out-bound notifications of the sort that Tim is talking
about, because it doesn't (natively) have something that the MOM
pub-sub and store-forward systems can map onto, besides email. In this
case, the Web may well have to adapt to the enterprise rather than
vice-versa.
It's very tempting to think of the Web as "vastly superior" to those
crufty old enterprise systems. Having been around dinosaur wranglers
for a few years now, I'm not so sure: those mainframes and MOM systems
are not going anywhere because they *work* -- 24/7/365, worldwide, in
the face of all sorts of hardware and communications failures, years
and years after the outside world stopped paying them any attention.
It only took one "but, why don't you just use HTTP?" pratfall at a
mainframe shop for me to learn that lesson :-) I recommend keeping a
*very* open mind on this subject, because both the mainframe/MOM
architecture and the Web architecture have proven themselves very
robust in practice. It's not at all clear who can learn most from whom
when they come together, and in Tim's "syndicate account updates to
consumers over the Web" scenario, they definitely come together.
|
