>> slonik is for the most part a wrapper to the plpgsql functions within
>> the
>> database, so the idea of making slonc an interactive client is not as
>> far
>> fetched as you may first think.
>>
>
> But this is a disadvantage as well:
> slonik's dialect isn't very user-friendly regarding to an interactive
> language,
> you have to play with cluster id's, table id's and node id's and this
> makes it
> a little bit "rough"....btw. this bothers me in the altperl scripts as
> well

One of my ideas for Slony-I 1.1 had been to introduce the notion of "node
names."

I discovered that managing this was *way* more difficult than it appeared.
 Adding and managing names would introduce a great deal of complications.

Korry Douglas then sailed in as a "savior", to my mind, when he
contributed code for keywords 'include' and 'define'.  That allows having
symbolic names in scripts, albeit in a way that is invisible to the
cluster's direct configuration...

It seems to me that your "interactive slonc" will need to have a LOT of
internal mechanism in order to work.  You would need to fairly much define
a full-scale language with variables, operators for "string smashing," and
such like.  There are two directions that can take you:

1.  You wind up having designed a new scripting language with a REPL. 
<http://en.wikipedia.org/wiki/Read-eval-print_loop>  Seems like a big
mistake to me to start such a project from scratch.

2.  You embed Slony-I functions into some existing REPL-based scripting
language.

Which more or less implies embedding it into Python or into some flavour
of Lisp.

I'd be interested to see that; I'm not sure that either option will be
forcibly considered "user friendly."

I'm not saying "don't bother trying," but it would be a mistake to take
the project on without a reasonably clear idea of the scope...

_______________________________________________
Slony1-general mailing list
[email protected]
http://gborg.postgresql.org/mailman/listinfo/slony1-general

Reply via email to