Talk:Telescript (programming language)

Implementation
The Telescript language may have defined all these cool things, but there is only one implementation that I ever knew about: AT&T's PersonaLink system. And the implementation sucked hard. It could not marshal code, and barely execution state. The same exact code had to be running within the device, and within the Place on the server. Some information from the client device could be serialized over the 2400 bps modem (UGH), and passed to an instance of the agent within the Place on the server. Then it would "do something" and then marshal a response back. The server infrastructure was so bloated and horrendous, that it could only execute about FIVE agents at one time. Needless to say, this system had no possible way to operate "in the wild" with that little scale.

The concept of agents flying around and performing actions and collecting data is a total myth. Academic paperwork that was never realized in implementation.

Worse: the marshalling was effectively, "strip everything out, so we only send the barest minimum of data". In essence, we used an agent to ship over a "packet" of binary data, as if we were operating a normal client/server modem connection. There were zero advantages to using Telescript and agents in this scenario.

We were pissed because General Magic told us "use agents", rather than sticking to normal modem communications. So we got saddled with the agent paradigm (which we worked around), and with the PersonaLink servers (which never succeeded). The whole implementation was crappy.

Gstein (talk) 02:27, 26 March 2016 (UTC)


 * , do you know of AT&T ever published a post-project roundup? It would make an excellent addition to this article. Maury Markowitz (talk) 14:11, 28 October 2016 (UTC)


 * ... sorry, I don't. We dropped all development when the platform collapsed, and moved to web-based commerce. Can't say that I looked back :-P Gstein (talk) 04:16, 22 December 2016 (UTC)