[ofa-general] [PATCH] libibverbs: Added the man page verbs.7

[Hal Rosenstock]

On Tue, 2008-02-26 at 09:27 +0200, Dotan Barak wrote:
> Hal Rosenstock wrote:
> > On Mon, 2008-02-25 at 13:19 +0200, Dotan Barak wrote:
> >   
> >> Or Gerlitz wrote:
> >>     
> >>> Sean Hefty wrote:
> >>>       
> >>>> The verbs also support iWarp devices and are not necessarily 
> >>>> restricted to the
> >>>> 1.2 IB spec definitions.  It might make sense to state that the IB
> >>>> implementation is based on the 1.2 spec in an IB specific section, 
> >>>> but keep the
> >>>> general documentation transport neutral at this point.
> >>>>         
> >>> Sure, the page would be changed to reflect that.
> >>>
> >>> Or.
> >>>       
> >> Sorry, i didn't find the time to get to it until now.
> >>
> >> I changed the problematic sentence to:
> >> "This library is an implementation of the verbs based on the Infiniband 
> >> specification volume 1.2 chapter 11."
> >>     
> >                 ^^^^^^^^^^
> >                 volume 1
> >
> > Also, should this refer to IBA 1.2.1 rather than 1.2 (if that was what
> > was intended by the 1.2 reference) ?
> >   
> I must admit that i didn't have a chance to check IB spec 1.2.1 BUT
> 1) Many of the features are implemented by the IB devices, and i don't 
> know if all of them
>      behave according to spec 1.2.1

Aren't any of those changes optional so I'm not sure I see the issue
here ?

> 2) Spec 1.2 introduced new features (such as SRQ and some more) which 
> didn't exist in
>     spec 1.1, so i wanted to make to make sure that they are supported. 

Sure; there are spec changes going forward but the spec is backward
compatible (and new features are optional).

> (as much as i know,
>     most of the users don't have a copy of the IB spec )

The spec has been publically available for quite a while now (not just
available to IBTA members).

> 3) If one month from now, spec 1.2.2 will be published; should we update 
> this file?

That's actually a larger question affecting more than this just this
file.

One approach would be to indicate the latest spec supported at the time
of release.

> I think that sentence is good (spec 1.2) for now...

I'm not sure about its "goodness". In this particular place, there may
not be much harm either way but in others it is misleading and
inaccurate.

> Do you think that we should remove the spec version completely?
> (i don't think that it is wise to update the version unless there is a 
> good reason for it ..)

That gets rid of the overhead of dealing with maintaining the spec
version. However, eliminating the spec version leaves the version open
which is not a good thing in all cases.

-- Hal

> thanks
> Dotan