Re: [gentoo-dev] maybe it is time to put portage queries into a database.

[Tony Clark]

On Wednesday 25 June 2003 00.16, Brian Jackson wrote:
> Totally agree. I have 5 Gentoo boxes at work (probably more soon), and it
> takes a ridiculously long amount of time to do searches. One thing I would
> suggest is to develop as an external tool, and once it matures, then shoot
> for inclusion into portage. another suggestion would be for it to support
> more than one type of db (sleepycat, postgres, mysql, gdbm for example). If

Yes, it would be nice to make it that way.  I guess something like client -> 
client-db-interface-layer -> db-of-choice.

> you try to make some postgres zealot put mysql on just to query packages,
> you are going to get some very unhappy words directed your way. I can try
> to help also in my spare time.

Well there are plenty of perl modules to handle client-db-interface-layer but 
I suspect they will all have different APIs for the client.  My thoughts are 
get one system to work then build on that.  So long as the design is modular, 
that should be too difficult for those interested. :)

tony

>
> --Brian Jackson
>
> On Tuesday 24 June 2003 05:00 pm, Tony Clark wrote:
> > I wanted to quickly find a package with emerge -s whatever.  I noticed
> > that it seemed to be taking a long time now for basic searches.  A couple
> > of random example are presented below.  Now in my case on the machine in
> > question, it is probably a bit slower than need be as /usr/portage is nfs
> > mounted but neither the nfs server, this client are what would be called
> > slow machines and I run 100baset ethernet.
> >
> > 12 months ago I thought the search times where acceptable, nowdays they
> > are pushing it and with the number of packages going into portage sooner
> > rather than later it is going to be unacceptal.  I would also suggest a
> > centralised server approach using something like mysql as apposed to
> > localised databases based on berekley or something else.  reason being
> > that a number a ppl are running home lans and they are increasing in
> > popularity and there is no need to store N machines worth of data. mysql
> > comes to mind as it is a popular database as there is a good chance that
> > for some other application it is already running on the network.  I know
> > in my case thats so.
> >
> > I know this has been discussed in the past but always put down for
> > reasons I largely agreed with, but the number of packages and times have
> > changed. I don't mind helping out with this with some perl etc.  What I
> > was thinking on for the initial implementation would be just to store the
> > information required for searching.  I don't see a need to store all
> > dependcies etc so the installation process actually becomes database
> > driven.  In this case the queriey times are so short to the actually
> > installation time, no noticable speed up will take place.
> >
> > The floor is open, let the flames begin :)
> >
> > tony
>
> <snipped times>

-- 
Contract ASIC and FPGA design.
Telephone [phone]
http://pgp.mit.edu:11371/pks/lookup?op=vindex&search=0x633E2623



--
gentoo-dev@g.o mailing list