Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Is it time for the Winboard Protocol to go the way of the Dodo?

Author: Vasik Rajlich

Date: 02:33:45 03/11/05

Go up one level in this thread


On March 10, 2005 at 18:42:53, Lance Perkins wrote:

>They are meant to illustrate that efficiency is an attribute of protocol design.
> The move list feature of UCI is obviously inefficient compared to what is
>already in xboard and ICS. Whether or not the inefficiency of UCI is acceptable
>is another matter - a debatable one - especially after it is has become obvious
>that the statement "UCI is not a network protocol" is not exactly true. Why
>invent an inefficient protocol to replace an efficient one? Hmmm.
>

There are many examples in computer science where we accept inefficiency for the
sake of elegance and programmer productivity.

In some cases (though not here) the inefficiency is considerable - for example,
interpreted/JIT-compiled languages.

Vas

>---
>
>On March 10, 2005 at 16:02:56, Anthony Cozzie wrote:
>
>>On March 10, 2005 at 14:29:21, Lance Perkins wrote:
>>
>>>On the contraty, this is not about bandwidth. Think 'protocol' design.
>>>
>>>API's is another form of protocol. When you write your Search function requring
>>>the alpha and beta values, do you pass these 'integers' as 'strings' and then
>>>convert them to integers inside the function? The values are naturally integer.
>>>Why represent them as strings.
>>
>>Strawman #1.  What does this have to do with anything?
>>
>>>Since you insist to make this a bandwidth issue, what makes you think that a
>>>chess engine protocol will not be used over the network? The ThinkerBoard
>>>package comes with a utility called RemoteThinker/RelayThinker that allows a GUI
>>>to run from one machine and the engine to run from another machine. Your GUI
>>>will have no clue that engine is actually remote.
>>
>>Strawman #2.  What does this have to do with anything?  I already proved with
>>hard numbers that the bandwidth difference is negligible.
>>
>>>In protocol design, when you invoke a service, you should be transport-safe. The
>>>service can be on the same machine or it can be on another machine.
>>
>>Strawman #3.  What does this have to do with anything?  Clearly any protocol
>>that goes over a pipe can go over a TCP stream.
>>
>>>These are very basic computer science concepts.
>>
>>And horribly misapplied.  If I wanted to read CS 101 again, I would.  You have
>>exactly two sentences worth considering: "Engines are naturally stateful. Why
>>invent a protocol that treats them differently."  Which boils down to "I don't
>>like UCI".  Which is fine, but not exactly the most logical of arguments.
>>
>>anthony



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.