[FoRK] mod_pubsub way ahead of its time: Comet solves Ajax-envy
Damien Morton <
fork at bitfurnace.com
> on >
Wed Mar 8 19:07:25 PST 2006
> XMLHTTP is almost completely irrelevant. It was only important in that
> users had a gee-whiz payoff from the accomplishments of the client-side
> hackers. Those accomplishments came from somewhere else -- bringing the
> discipline of FOSS development over from Gnu/Linux to apps hosted in the
A grand claim. In fact, in my case, it came from my experiences with the
capabilities of Macromedia Flash, and a general 'rich-client'
development mentality that was irked at the inefficient page-at-a-time
nature of the web. At that time, the only UNIX/FOSS exposure I had was
having to use UNIX at college, and a liking for Python.
> In effect the browser hosted running environment was just another fucked
> up flavor of unix to wrapper and turn into something orderly, albeit one
> without POSIX.
Ahem. Bullshit. This has nothing to do with unix, and everything to do
with subverting the W3C page-at-a-time model of the user experience,
transforming it into a rich-client model (which has traditionally been
the domain of Microsoft).
Probably the first major application I ever saw using this technique was
the MSDN website, which dynamically loaded parts of a treeview in the
in the main page. That was back in 1999, maybe even earlier.
I think you're really reaching to make AJAX (I hate that term) a
UNIX/FOSS thing. Certainly, it gives the UNIX/FOSS community a way to
present cobbled-together rich-client applications to predominantly
windows users, but its going to be extremely hard to beat a
purpose-built rich-client browsing environment (such as the one Microsft
is currently prepping in the form of XAML, whose dominance is far from
assured, but whose capabilities far outstrip HTML with or without AJAX).
On the subject of subverting the W3C page-at-a-time model, I think the
AJAX community is going to hit the same problems that the Flash
community has been striking for years; incompatibility with search
engines and an anti-functional or inconsistent back-button model. For
some reason, people just love their back-button, but its almost
impossible to come up with a consistent meaning for it in a rich-client
> On Thu, 9 Mar 2006, Damien Morton wrote:
>> Well, yes all of those and more. Dynamically loaded scripts were also
>> used, as were dynamically loaded data in the style of JSON.
>>> I think the magic ingredient was the IE ActiveX component (XMLHTTP),
>>> which was then (re)implemented in Mozilla/Firefox. Before that, you
>>> had to use stone knives, bear skins, and hidden IFRAMEs.
More information about the FoRK