From mboxrd@z Thu Jan  1 00:00:00 1970
From: Dirk Bergstrom <dirk@juniper.net>
To: 'Peter Novodvorsky' <nidd@altlinux.ru>, gnats-devel@sources.redhat.com
Subject: RE: modular database backends
Date: Tue, 29 May 2001 00:11:00 -0000
Message-id: <C0D6C1C24CDBE1449BFEF1B72AFBF3A709A5AA5E@postal.jnpr.net>
X-SW-Source: 2001-q2/msg00119.html

a database backend for gnats, huh?  i've been thinking about this myself.
in fact, it's quite likely to be a major project for me this year.  i
haven't gotten too far along, but i will soon have a lot more time to devote
to thinking about it.  i wouldn't describe myself as a high powered C
programmer (cough cough...), but i've got clue, time, and energy.

we should gather together interested parties, and kick some ideas around.

--
Dirk Bergstrom              dirk@juniper.net
____________________________________________
Juniper Networks Inc.,  Engineering Web Guru
Tel: 408.745.3182          Fax: 408.745.8905


> -----Original Message-----
> From: Peter Novodvorsky [ mailto:nidd@altlinux.ru ]
> Sent: Sunday, May 27, 2001 11:18 AM
> To: gnats-devel@sources.redhat.com
> Subject: modular database backends
> 
> 
> 
> Hello!
> 
>         I always thought about BTS as database and tools to work with
> it. And these things are different. In gnats I see opposite 
> thing: database,
> tools and servers share one source and tools sometimes do too much.
> 
>         So I've began work on making database backends to 
> gnats modular. 
> Just now, gnats has two "backends", network (gnatsd) and filsystem. If
> we make system of backends modular we could have SQL backend, 
> etc., etc.
> 
>         Second thing I don't like in  GNATS, it doesn't has ACLs. This
> modular framework should support them.  Here is the header file that
> will briefly describe what I mean:
> 
> struct _gnats_database {
>   URI *U;
>   
>   FILE *serv_read;
>   FILE *serv_write;
> 
>   void *module_handle;
> 
>   int (* init) (ErrorDesc *err, gnats_database);
>   int (* add_pr) (ErrorDesc *err, gnats_database, PR);
>   int (* update_pr) (ErrorDesc *err, gnats_database, PR);
>   int (* query_pr) (ErrorDesc *err, gnats_database, QueryExpr);
>   int (* disconnect) (ErrorDesc *err, gnats_database);
> };
> 
> typedef struct _gnats_database *gnats_database;
> 
> gnats_database gnats_new_database (ErrorDesc *err, URI *U);
> int gnats_destroy_database (gnats_database db);
> 
>         module_handle is object handle that is returned by dlopen.
> URI is:
> 
> struct _URI
> {
>   char *scheme;
>   char *host;
>   char *user;
>   char *pwd;
>   char *path;
>   unsigned int port;
> };
> 
> typedef struct _URI URI;
> 
>         And there is a function
> 
> URI *gnats_parse_uri (const char *gnats_uri, int length);
> 
> that takes URI (for example 
> gnatsd4x://nidd:pwd@nidd.debian.net:1529/database_name) and converts
> it in such structure (gnatsd4x is 4.0 gnatsd protocol).
> 
>         I'm currently implementing gnatsd4x backend. After this I'll
> make query-pr, pr-edit, gnatsd and other tools to work with 
> this model.
> And then I'll begin to make SQL backend.
> 
>         If everything will be OK in the end following will be 
> possible:
> 
>            user -> client -> gnatsd -> SQL server (MySQL, Postgres)
> 
>         What do you think of this crazy idea?
> 
>                                                 Thanks,
> -- 
> Peter Novodvorsky     http://www.altlinux.ru/    AltLinux Team, Russia
> Debian.Org                                       
http://debian.org/~nidd
            Debian  ---  no need to  wait for tomorrow.