Re: FCIP: Comment 120 - connection endpoint

[Ralph Weber <ralphoweber@compuserve.com>]

Mallikarjun,

I think we should stop worrying about what the FCIP Entity
'is' and start focusing on what it 'does'. One could argue
that taking this approach fails to address the clarity issue
you have described. But, I feel that it gets the names out
of the way and directs attention to the requirements.

With this in mind, I propose changing the sentence below
figure 4 that we have been discussing as follows.

From:

The FCIP Entity is the connection interface point for the
IP Network and is the owner of the IP Address and Well Known
Port used to form TCP Connections.

to:

The FCIP Entity receives TCP connect requests on behalf of
the FCIP_LEPs that it manages. In support of this, the FCIP
Entity is the sole owner of at least one TCP port/IP Address
combination used to form TCP Connections. The TCP port may be
the FCIP well known port at a given IP Address.

Thanks.

.Ralph

"Mallikarjun C." wrote:

>
> Ralph,
>
> > FCIP appears to be a case where everything you have ever seen is turned
> > on its head.
>
> Actually, the reality isn't so bad - on-the-fly instantiation of application structures
> is a fact of life for most TCP listeners, for ex., iSCSI session on targets.
>
> I think the draft is simply describing the notion of "connection interface point" from TCP's
> perspective, while describing the idea of "connection endpoint" from application (FCIP)
> perspective.
>
> Perhaps this distinction needs to made clear in description of these ideas -
> at least this wasn't obvious to me earlier.
>
> Regards.
> --
> Mallikarjun
>
> Mallikarjun Chadalapaka
> Networked Storage Architecture
> Network Storage Solutions
> Hewlett-Packard MS 5668
> Roseville CA 95747
> cbm@rose.hp.com
>
> ----- Original Message -----
> From: "Ralph Weber" <ralphoweber@compuserve.com>
> To: <ips@ece.cmu.edu>
> Cc: "Mallikarjun C." <cbm@rose.hp.com>
> Sent: Friday, May 10, 2002 7:57 PM
> Subject: Re: FCIP: Comment 120 - connection endpoint
>
> > Mallikarjun,
> >
> > It looks like we are between a rock and a hard place here.
> >
> > > ... In all the implementations I had seen, your connection endpoint
> > > is where you sent the connection requests to - i.e. the connection
> > > interface point.....
> >
> >
> > In FCIP, the FCIP Entity is the place to which TCP connect requests
> > are sent. When the TCP connect request arrives, the FCIP_DE does not
> > exist, meaning that FCIP cannot have the TCP connect requests being
> > directed to the FCIP_DE because it is flat out not there.
> >
> > Only after a TCP connect request arrives and is validated (FSF exchange,
> > and possibly ASF exchange) does the FCIP_DE get created at tied to the
> > endpoint of the newly established TCP Connection.
> >
> > Thus there is a very real difference between the TCP endpoint (which
> > is connected to the FCIP_DE) and the connection interface point
> > (which is inside the FCIP Entity).
> >
> > Short of a serious FCIP rewrite (probably with major confusion added),
> > I do not see any way around this critical distinction.
> >
> > Sorry.
> >
> > .Ralph
> >
> >
> > > > Consider the revised sentence again:
> > > >
> > > >   "The FCIP Entity is the connection interface point for the IP Network
> > > >   and is the sole owner of at least one TCP port/IP Address combination
> > > >   used to form TCP Connections. The TCP port may be the FCIP well
> > > >   known  port at a given IP Address."
> > >
> > > This is good (but see below).
> > >
> > > >
> > > > Viewed in context, "connection interface point" is synonymous with
> > > > "the place to which TCP connect requests are directed". It is not
> > > > and is not intended to be synonymous with the endpoint of a TCP
> > > > Connection once that TCP connection is formed.
> > >
> > > It's unclear to me how the two are different.  In all the implementations I
> > > had seen, your connection endpoint is where you sent the connection
> > > requests to - i.e. the connection interface point.....
> > > --
> > > Mallikarjun
> > >
> > > Mallikarjun Chadalapaka
> > > Networked Storage Architecture
> > > Network Storage Solutions
> > > Hewlett-Packard MS 5668
> > > Roseville CA 95747
> > > cbm@rose.hp.com
> >
> >