[openib-general] RE: [PATCH 2/4] opensm: remove unused osm_pkey_mgr_t object

Eitan Zahavi eitan at mellanox.co.il
Wed May 3 23:01:17 PDT 2006


Hi Sasha,

I think changing the basic concept of a "manager" in OpenSM is not just
a cleanup issue.
I am for improving the code - but not for breaking its basic
architecture.

If you find dead code or unused code - let's fix it. But please try to
keep the "structure" untouched. 

I have a many ideas for how OpenSM could be re-written in a better way
too.
(Like avoiding SA code duplication by using C++ or C virtual functions)
but I do not think it is a small change - but rather a big one (actually
a re-write). 
One day we might decide a re-write of the SM is required but this should
not be taken lightly as it would probably take a significant effort and
a few years to get back to the current status.

Eitan

Eitan Zahavi
Design Technology Director
Mellanox Technologies LTD
Tel:+972-4-9097208
Fax:+972-4-9593245
P.O. Box 586 Yokneam 20692 ISRAEL


> -----Original Message-----
> From: Sasha Khapyorsky [mailto:sashak at voltaire.com]
> Sent: Thursday, May 04, 2006 3:26 AM
> To: Eitan Zahavi
> Cc: Hal Rosenstock; openib-general at openib.org; Ofer Gigi; Yael Kalka
> Subject: Re: [PATCH 2/4] opensm: remove unused osm_pkey_mgr_t object
> 
> Hello Eitan,
> 
> On 11:41 Tue 02 May     , Eitan Zahavi wrote:
> >
> > I really do not like this patch. I think that although it does not
break
> > the code TODAY, it will be reversed later.
> > OpenSM uses the concept of "manager" for each of the algorithms
used.
> > One could claim that all these managers are redundant and could be
> > replaced by an extension to the osm object. This is true but will
result
> > with a non clear boundary between the managers.
> 
> "manager" concept is fine, but I don't see how useless structure
should
> help in implementing this. OTOH there are tons of duplications and
> unnecessary code in OpenSM today - we need to improve this.
> 
> > Although there is no right or wrong on this kind of issues, I think
that
> > the winning argument is that today OpenSM is written according to
the
> > above simple rule.
> 
> Hmm, so what is your argument - "it is so now, don't change it"? But
the
> goal is to improve things.
> 
> Sasha.



More information about the general mailing list