Re: On the Go Modem Protocol

[David Fotland <fotland@xxxxxxxxxxxxxxx>]

Optional features are bad in general.  People will implement
different subsets, and the programs won't be able to
communicate.

The problem with any new protocol is comvincing anyone to
spend time implementing it :)  If you plan to make it easy
by making a DLL or other simple interface to the protocl, then
you should probably do that interface first to an existing
protocol like GMP or IGS, so you have a base of programs
to play against.

I don't know why people don't like IGS.  Many Faces and Gungo already
play there.

David

At 10:16 AM 5/16/99 +0300, Antti Huima wrote:
>I think it would be nice if GMP could be replaced with something with at
>least the following features (some of these are present in GMP):
>
>1. Based on human-readable encoding, such as HTTP, SMTP etc.
>2. No error correction, assume underlying layer handles it.
>3. No small limits on values transmitted.
>4. Modular structure: there is the core protocol and then optional
>   features that can be supported but not necessarily. Such features
>   could include:
>   1. Board setup (for testing a go program)
>   2. Overriding opponent's moves (for testing a go program)
>   3. Method for transferring markup from a go client to a protocol
>      client (go programs could use this for annotating board positions
>      and `explaining' their moves --- single protocol client would
>      suffice)
>   4. end-to-end conversation (`talk')
>   X. proprietary extensions
>5. No conflict situations: the protocol must be as deterministic as
>   possible.
>6. No timeouts, except for those induced by the underlying carrier.
>7. Support for multiple character sets.
>
>I don't consider the IGS protocol to be a good choice at the moment,
>although my opinion is not based on extensive study.
>
>-- 
>Antti Huima
>SSH Communications Security Oy
>
>
>
>