[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: sector alignment for DataOut PDUs?
>>>>> "Rod" == Rod Harrison <firstname.lastname@example.org> writes: Rod> Paul, I strongly object to your alternative of tightening the Rod> spec wording to force all PDUs to be max sized except the Rod> last. As I mentioned in an earlier email, since draft v09 we Rod> don't have a negotiated PDU size we have arbitrary maximums set Rod> by the receivers. Forcing an initiator to send "full" PDUs is Rod> unreasonable since the target may set a huge MaxRecvPDULength. Rod> The intent of the current wording was, if I remember correctly, Rod> to have the initiator always send as much unsolicited data as Rod> possible. This was to avoid the target having to wait for the F Rod> bit on the unsolicited data before issuing R2T(s) for the Rod> remaining payload. Note that unsolicited data, and its maximum Rod> size, are negotiated parameters and that that current SHOULD is Rod> quite different from forcing initiators to always send "full" Rod> DATA-OUT PDUs. That's a good point. So for unsolicited data we have a negotiated max length, but you're allowed to send less than that. Similarly, for R2T the target requests a length and the initiator is allowed to send less than that. That can be tightened up, but doing so doesn't constrain the individual PDUs that are being sent, only the whole burst. I was concerned with the individual PDUs, so a different rule is needed there. And as you point out, the max PDU length is no longer suitable for that. (Undoing the change of what Max PDU length means back to what it was in -08 doesn't sound like the right way to address this.) Rod> Please note that I don't object to your original proposal, just Rod> to the alternative which I believe would be a huge functional Rod> change. Thanks... your comments make sense to me. paul
Last updated: Wed Feb 27 15:18:07 2002
8911 messages in chronological order