Re: Towards Consensus on TCP connections

[julian_satran@il.ibm.com]

That is a very good point. We too avoided touching structures we didn't
have too (this is how we ended-up with huge LUNs). Bridging will be simple
for "proper networks" but might
involve more logic for mapping to SCSI-2 or very simple SPI. And I would
appreciate any more insights into bridging.

We  assume that drivers for widely available OSs will be readily available
in source.

Julo


Stephen Bailey <steph@cs.uchicago.edu> on 17/08/2000 18:02:06

Please respond to Stephen Bailey <steph@cs.uchicago.edu>

To:   ips@ece.cmu.edu
cc:    (bcc: Julian Satran/Haifa/IBM)
Subject:  Re: Towards Consensus on TCP connections




> It is also very important to note that SPI-4, FCP-2, and SVP are
> active projects that share a goal of common packet formats to
> simplify implementation of bridges between the parallel bus,
> fibre channel, and VI.

SST also has this goal.  The addressing, and command and status
formats are as similar as we could make them to FCP.  It was quite
handy to do this.

On one hand, given the mismatch in other aspects than information unit
format between iSCSI and FCP (specifically, the congestion avoidance
requirements), cribbing IU format from FCP[-2] is not going to go a
huge distance towards making iSCSI easier to implement.  On the other
hand, making everything possible about iSCSI `same as before', WILL
still help substantially to anybody approaching the technology (device
implementer, driver writer, etc.).  For example, in SST, it was useful
to be able to start with an FCP driver source and just adapt that to
fit the ST layer interface.  Again, not a huge deal, but it does help
make it seem like a familiar task, rather than a bizarre one.

Steph