Date: Wed, 26 Mar 1997 14:34:16 -0500
From: Robert Jasiek <jasiek@berlin.snafu.de>
Subject: Re: Why we need FF[4] [Was: SGF - some answers] 

> The pass move is the biggest issue for me forcing me to write FF[3], but
> there are others like: enforcing the format of game info properties,
> arrows and lines, and the fact that we don't have a final spec and I
> am finishing my sgf code this week.

The last point is important, but finally (after a year or so) you will
probably update your program and then use the new specification. AR and LN
are no problem, you just ignore and do not use them. Do you intend to 
continue [tt] even after your next program revision? Else Anders' compromise
could be used. What exactly are the differences concerning the format of
game info properties, why can you not use FF[4] for them? (I am not familiar 
with Ishi-press conversion trouble.)

> I decided to read all triples, but convert them to comments.  MFGO's editor
> won't allow you to set a triple, and it won't write triples.  This avoids
> the cryptic symbols, and is easy to convert to Ishi format.

This is reasonable.

> Please lets not have the metalanguage argument again.  I guess I can see
> that Europeans want language independent symbols, and automatic (even if
> bad) translation for multipl languages.  But my product is English, or
> will be localized separately to each European language.  I'd rather give the
> user clear, correct text in his/her native language, than force him to
> learn some symbology (like the difference between !? and ?!).
> -David

Sorry, you have not understood the ML concept. Maybe you stopped reading
when the discussion was about cryptic symbols. Now there are none. An ML
property does not create problems for you. You may simply ignore it. Or
you may treat it like triples and output the text as C text. You do not
need user input, user interface, or even translation. This may be left
for other viewers. Any ML text should be stored in English. Even if not
there is no difference to foreign language in a C property. ML is not
worse than a triple or often a C text which do not use good English, 
either.

I would like to hear your opinions:
- Who needs the doubles' properties?
- Who cannot tolerate an additional ML property (using text as property   
  value) by just skipping it?
- Who likes ML to replace or complete the doubles?

ML can be included in the property list for two years to see whether it is
used. No one needs to care for handling of ML text other than simple
skipping or C text conversion. I do not mind to write a user guide for
interested programmers about advanced text handling. BTW, ML replacing
the doubles would abandon all their trouble and greatly simplify the
property specification list.

-- Robert Jasiek   jasiek@berlin.snafu.de