Ok,
I get impatient now. So I will quote too.
I just learend that this will add an awfull lot of dead text
> to the index and it seems not viable to handle it that way.
>
As a user, I use what I write. So it's not dead, even if it's slow. As a
user, I do whatever I want. Among other things, I want to classify my games
according to projects.
> [Category "007030123803"]
>
> And resolve the full strings from
>
> 007 Castle Destruction
> 030 Isolated Pawns for the Advanced Player
> 123 Rook Lift
> 803 Double Bishop Sacrifice
>
> by some table.
>
I understand the need to have normalized themes. But really, the problem of
normalizing themes is as difficutl as the players' and event's names. A
bigger problem is the one the solution presupposes : what is the ontology of
chess themes ? Another bigger problem is : how do the concepts of this
ontology relates to games ?
Saying that Isolated Pawns for the Advanced Player should be a predefined,
numbered category is beyond me. If I want to write an article about
creative solutions to strategical problems in some games played by Topalov
in the 1990s, what number should we give to the category ? I only wanted to
say that a Category is also useful as something not predefined. It could
denote projects, like the example I gave. The concept of category, unless
tied to a formalized ontology, is like Goodman said about similarity : a
pretender, an impostor, a quack, unless we can become more explicit about
how it relates to the interests of the users.
Another thing I wanted to say is that you can attach more than one category
to a game. In a technical sense, it does not sound like a category. If you
had one, you don't have to say
rook-endings/techniques/lucena-bridge
because saying "lucena-bridge" is strongly typed, i.e. you can tell exactly
where it falls into your terminology without having to recall the parents.
Talking of categories means you have an ontology.
> You actually suggest chains of keywords. This is funny,
> from my professional point of view. It is said, that
> nobody uses this kind of deep indexing. (I do not believe
> that, but that's the result of quesioning the users.)
Keywords are here to stay. Just look at del.icio.us, youtube, last.fm, and
gmail, or just bibliographic references. I am no fan of keywords, as you
can guess from what I said earlier. But I can respect its popularity.
I'd suggest (oh, I suggest to much) to just start with some
> tournament and to fill a DB and see which of all those
> suggesions made so far are viable, what is possible at a
> resonable workload and so on.
I'd say that as long as we can't specify the scope of a predicate, be it a
category, a flag or a simple keyword, we're toast. We would never know to
what really the predicate attaches itself to. It looks like doing medieval
logic even when we now have quantificational theory. We
> Why? Your indexing is perfect. We do not do this in
> libraries as if we would do such a deep indexing people like
> me would have to be employed by the thousands in each
> discipline and we wouldn't do any other work than indexing
> literature at all. This is a bit expensive and maybe also
> way beyond the users needs. Be carefull here not to set up a
> to complex scheme with to ambitous targets.
>
How stupid what the users do should not be of our concern. The scheme is
very simple : keywords and two levels of titles. It suffices to write
anything that is short enough to be read in a life time. It does not need
any kind of modification from the application : it's just a practice to
promote.
> Wouldn't such stuff be more suitable within the games
> comments?
That's very possible. But as I said, it's a way to separate authorship.
Suppose a student has been given exercises, and messes up _his_ comments.
If he strips comments, he won't strip the question. I can think of more
important reasons. It's just a side issue anyway, something that quoting
never conveys, since it put the focus on everything it responds to.
> Wouldn't it be worthwile to think about storing it
> outside the DB?
If we don't have Ids, that won't do. Besides, I use the database to
generate stuff outside the base, not the other way around. That means I
want the application to extract the games and the positions I underlined
somehow. It can very well mean it outputs a table like you said. It could
very well be a report, with links to the games and positions.
> Thinking of Pascals comments about the index
> growing in astronomical numbers.
I don't see the step where the task gets polynomial. So if we have a
resource problem, it will arise as soon as we want one big database of
everything. The database of everything has the same problems with only key.
> Wouldn't it be better to
> say: Ok, I've a table that generates the trainings DB. This
> table contains the Game ID, the Ply where the quesion
> arrises, the question as such and the answer, just in
> separate coloumns?
I like this idea : we have a main database, from which we can generate
little, specific databases. And for those database, we could produce
something like Training Reports. I think what you're hinting at could very
well be reports, if we accept the idea that reports are views.
> Wouldn't it be a better approach to
> assemble the trainings DB then from the reference DB and
> adding this additional info from this table?
>
Yes, if we can link to positions. Now, we can't even link to games, so
we're far from it.
>
> 3. To the most interesting flags of Pascal's list, I would have opening,
>> so we could have only four flags :
>>
>> - Opening
>> - Middlegame
>> - Endgame
>> - Tactics
>>
>
> Ahm, do not reduce the 16 that are there. They're barely
> enough anyway ;)
Most of them are user-centered, so useless for CentriScid. They sure can
stay, but don't mean anything clear for me. So I don't flag much. I don't
know how White and Black Opening can translate from user to user, so they're
just tags, i.e. personal stuff. I never really cared about the concept of
novelty, which presupposes we already have access to every game in one
database. I would never use Queenside and Kingside, and it omits the
center. Pawn structure might be interesting, but if the structure is
unspecified, I wonder what it conveys other than middlegame : center ?
Brilliancy games can contain blunders too. User defined is a very strange
flag : it may be compared to the Star in Gmail, which I use to mark what I
read and want to keep. But since every flag are user-centric, I wonder what
that conveys.
***
To sum up :
- CentriScid could use flags that relates to overall games ;
- CentriScid could serve to generate specialized dbs and reports ;
- Keywords should be left to users ;
- Most of actual flags too ;
- Giving sample uses and reporting user cases are always necessary, and
sometimes sufficient.
-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Scid-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scid-users
