Re: Clause 4 comments
"Shah, Hemal" <[email protected]>
| Newsgroups | gmane.ietf.rddp |
|---|---|
| Message-ID | <[email protected]> |
Barry, Below is my response to editorial comments on CLAUSE 4 that you posted to the reflector in late Nov. Hemal * To: "RDDP" <[email protected]> * Subject: [rddp] Clause 4 comments to draft-ietf-rddp-ddp-01 * From: "Barry Reinhold" <[email protected]> * Date: Mon, 17 Nov 2003 13:33:37 -0500 * Importance: Normal * List-help: <mailto:[email protected]?subject=help> * List-id: IETF Remote Direct Data Placement (rddp) WG <rddp.ietf.org> * List-post: <mailto:[email protected]> * List-subscribe: <https://www1.ietf.org/mailman/listinfo/rddp>,<mailto:rddp-request@ietf. org?subject=subscribe> * List-unsubscribe: <https://www1.ietf.org/mailman/listinfo/rddp>,<mailto:rddp-request@ietf. org?subject=unsubscribe> * Sender: [email protected] _____ CLAUSE 4 Below are comments on clause 4. There are 32 editorial comments, most are nits, but there are some rewordings that are significantly different. Most of the rewording was made in an effort to improve the clarity of the concept being defined. None of the rewording is believed to be a technical change. 1, Editorial, Low, 4.1 under Advertisement Is: the act Suggest: The act Change is: Capitalize "the" [hvs] Change accepted. 2, Editorial, low, 4.1 under Data Delivery Is: Delivery is defined as the process of informing the ULP or consumer that a particular Message is available for use. Suggest: Delivery is defined as the process of informing the ULP or consumer that a particular message is available for use. Change: capitalization of "message" removed [hvs] Change accepted. 3, Editorial, low, 4.1 under Data Delivery Is: This is specifically different from "Placement", which may generally occur in any order, while the order of "Delivery" is strictly defined. Suggest: Delivery is distinct from "placement". Placement of data into ULP buffers may occur in any order, whereas the order of message "delivery" to the ULP is strictly defined. Change: Reword [hvs] Change not accepted. I would rather keep the text as it is. 4, Editorial, low, 4.1 under Data Sink Is: Note that the Data Sink can be required to both send and receive RDMA/DDP Messages to transfer a data payload. Suggest: Note that the Data Sink can be required to both send and receive RDMA/DDP messages to transfer a data payload. Change: capitalization of "message" removed [hvs] Change not accepted. Stick with the convention. 5, Editorial, low, 4.1 under Data Sink Is: Note that the Data Source can be required to both send and receive RDMA/DDP Messages to transfer a data payload. Suggest: Note that the Data Source can be required to both send and receive RDMA/DDP messages to transfer a data payload. Change: capitalization of "message" removed [hvs] Change not accepted. Stick with the convention. 6, Editorial, low, 4.1 under Node Is: A computing device attached to one or more links of network. Suggest: A computing device attached to a network by one or more links. Change: Reword [hvs] Change not accepted. Added a before the network 7. Editorial, low, 4.1 under ULP Is: The ULP for RDMA/DDP is expected to be an OS, Application, adaptation layer, or proprietary device. Suggest: The ULP for RDMA/DDP is expected to be an OS, application, adaptation layer, or proprietary device. Change: capitalization of application removed [hvs] Change accepted. 8. Editorial, low 4.1 under ULP Message Is: the ULP data that is handed to a specific protocol layer for transmission. Suggest: The ULP data that is handed to a specific protocol layer for transmission. Change: Capitalize the first word in the sentence [hvs] Change accepted. 9. Editorial, low 4.1, 4.2 Consistency of sections - The current draft has three sub sections, General, LLP, and DDP. Given that framework it appears that there should be a sub section for ULP containing (ULP, ULP message, ULP payload, ULP buffer, and ULP message length). However, the way a glossary is used is to quickly identify the meaning of a term. Given the size of the glossary a single level alphabetical organization is probably better. Suggest: Removing sub sections in the glossary. [hvs] Change not accepted. 10. Editorial, medium, 4.2 & 4.3 under LLP Connection and LLP Stream, also relates to 4.3 DDP Stream. The definitions of stream and connection could be more precise. Currently defined are DDP stream, LLP stream, and LLP connection. Within the draft document connection is used five times without a LLP qualification (clause 3.2 page 6, clause 4.1 under local peer, remote peer, and LLP stream, clause 10.4), stream is used four times (8.2, 9.1 item 1, 9.1 item 1 under non-zero case, and 10.4 item 3) without qualification. Suggest: Defining DDP stream, and DDP connection in terms of concepts significant to DDP. Define LLP connection and LLP stream by reference to LLP protocol specifications. Possible wording below: DDP stream - A logical channel between a local peer and a remote peer in which the ordering of a sequence of DDP messages is maintained. There is no guarantee of message ordering when a sequence of DDP messages is sent over more then one stream. DDP connection - An association between two DDP peers that provides a context for DDP streams. DDP streams can not exist outside the context of a DDP connection. If supported by the LLP, more then one DDP stream may exist within a single DDP connection as in SCTP. LLP connection - A connection as defined by the lower layer protocol. See [SCTP], [TCP]. LLP stream - The logical channel provided by the LLP that allows the ordered delivery of DDP messages between DDP peers. For SCTP the LLP stream is the SCTP stream (see [SCTP]). For TCP the ordered delivery service is provided by the TCP connection (see [TCP]). Note that TCP supports no more than one DDP stream per TCP connection. [hvs] Change not accepted. The current definitions of LLP connection, LLP Stream, and DDP Stream are fine. I do see a need to have a DDP connection definition. 11. Editorial, medium, 4.2 under ULPDU Is: Upper Layer Protocol Data Unit. The data record defined by the layer above MPA. Suggest: Upper Layer Protocol Data Unit. The data record defined by the layer above DDP. Change: Replaced MPA with DDP. [hvs] Replaced MPA with LLP. 12. Editorial, low, 4.3 under Data Placement Is: Placement (Placement, Placed, Places) - For DDP, this term is specifically used to indicate the process of writing to a data buffer by a DDP implementation. DDP Segments carry Placement information, which may be used by the receiving DDP implementation to perform Data Placement of the DDP Segment ULP Payload. See "Data Delivery". Suggest: Placement (Placement, Placed, Places) - The process of writing the ULP payload of a DDP segment to the data buffer specified in the header of the DDP segment. Placement and delivery have specific and distinct meaning in DDP. See "Data Delivery" and "Direct Data Placement". Change: Reword and simplify, reference Direct Data Placement for additional details. [hvs] Added "Direct Data Placement" at the end. Other text is still the same. 13. Editorial, low, 4.3 under DDP Control Field Is: a fixed 8-bit field in the DDP Header. Suggested: A fixed 8-bit field in the DDP header. Change: Capitalize "a", remove capitalization of header [hvs] Change accepted. 14. Editorial, low, 4.3 under DDP Header Is: The header present in all DDP Segments. The DDP Header contains control and Placement fields that are used to define the final Placement location for the ULP Payload carried in a DDP Segment. Suggest: The header present in all DDP segments. The DDP header contains control and placement information. (See clause 6) Change: Reword and added reference. [hvs] Change not accepted. I am fine with the current text. 15. Editorial, low, 4.3 under DDP Message Is: A ULP defined unit of data interchange, which is subdivided into one or more DDP Segments. This segmentation may occur for a variety of reasons, including segmentation to respect the maximum segment size of the underlying transport protocol. Suggest: A ULP defined unit of data interchange, which is subdivided into one or more DDP segments. This segmentation may occur for a variety of reasons, including segmentation to respect the maximum segment size of the underlying transport protocol. Change: Changed capitalization of DDP Segments... (nice definition BTW) [hvs] Change not accepted. Sticking with the convention. 16. Editorial, low 4.3 under DDP Segment Is: The smallest unit of data transfer for the DDP protocol. It includes a DDP Header and ULP Payload (if present). A DDP Segment should be sized to fit within the Lower Layer Protocol MULPDU. Suggest: The smallest unit of data transfer for the DDP protocol. It includes a DDP header and ULP payload (if present). A DDP segment should be sized to fit within the lower layer protocol MULPDU. Change: Capitalization [hvs] Change not accepted. 17. Editorial, low 4.3 under Direct Data Placement Is: A mechanism whereby ULP data contained within DDP Segments may be Placed directly into its final destination in memory without processing of the ULP. This may occur even when the DDP Segments arrive out of order. Out of order Placement support may require the Data Sink to implement the LLP and DDP as one functional block. Suggest: The placement of the ULP payload in a DDP segment into its final memory location without processing by the ULP. This may occur even when the DDP Segments arrive out of order. Out of order placement support may require the data sink to implement the LLP and DDP as one functional block. Change: Reword and capitalization [hvs] Change not accepted. Rewording is not making it better. 18. Editorial, medium 4.3 under Direct Data Placement Protocol Is: Also, a wire protocol that supports Direct Data Placement by associating explicit memory buffer placement information with the LLP payload units. Suggest: A wire protocol that enables direct data placement by associating placement information with the ULP payload units. Change: Wording, replaced LLP with ULP. [hvs] Change not accepted. LLP is correct because placement information is in LLP payload. 19. Editorial, low 4.3 under Message Offset Is: For the DDP Untagged Buffer Model, specifies the offset, in octets, from the start of a DDP Message. Suggest: The offset, in octets, from the start of a DDP message. Used with the untagged buffer model. Change: Wording, capitalization [hvs] Rewording not accepted. Not seeing advantage of rewording. 20. Editorial, low 4.3 under Message Sequence Number (MSN) Is: For the DDP Untagged Buffer Model, specifies a sequence number that is increasing with each DDP Message. Suggest: A sequence number in the header of a DDP segment. Used with untagged buffers. See clause 6.3. Change: Replaced definition with reference to clause 6.3. The two did not appear to be consistent; goal is to have a single definition in spec. [hvs] Change not accepted. Keeping the definition unchanged. 21. Editorial, low 4.3 under Protection Domain Is: A Mechanism used to associate a DDP Stream and an STag. Under this mechanism, the use of an STag is valid on a DDP Stream if the STag has the same Protection Domain Identifier (PD ID) as the DDP Stream. Suggest: A Mechanism used to associate a DDP stream and an STag. Under this mechanism, the use of an STag is valid on a DDP stream if the STag has the same protection domain identifier (PD ID) as the DDP stream. Change: Capitalization [hvs] Change not accepted. 22. Editorial, low 4.3 under QN Is: For the DDP Untagged Buffer Model, identifies a destination Data Sink queue for a DDP Segment. Suggest: The destination data sink queue for a DDP segment. Used with the untagged buffer model. Change: wording, capitalization [hvs] Change not accepted. 23. Editorial, low 4.3 under Steering Tag Is: An identifier of a Tagged Buffer on a Node, valid as defined within a protocol specification. Suggest: An identifier of a tagged buffer on a node, valid as defined within a protocol specification. Change: capitalization [hvs] Change not accepted. Sticking with the convention. 24. Editorial, low 4.3 under Tagged Buffer Is: A buffer that is explicitly Advertised to the Remote Peer through exchange of an STag, Target Offset, and length. Suggest: A buffer that is explicitly advertised to the remote peer through exchange of an STag, target offset, and length. Change: Capitalization [hvs] Change not accepted. Sticking with the convention. Changed Target Offset to Tagged Offset just to be consistent in the doc. Thanks for pointing this out. 25. Editorial, medium, under Tagged Buffer Model Is: A DDP data transfer model used to transfer Tagged Buffers from the Local Peer to the Remote Peer. Suggest: A DDP data transfer model that allows the local peer to advertise a named buffer. (See clause 3.1) Change: Reword and added reference to more complete definition [hvs] Change not accepted. I will stick with the current definition. 26. Editorial, low 4.3 under Tagged DDP Message Is: A DDP Message that targets a Tagged Buffer. Suggest: A DDP message containing a Stag. Messages which contain Stags use the tagged buffer model. Change: Reworded to remove the word "targets" [hvs] Change not accepted. I will stick with the current definition. 27. Editorial, low, 4.3 under Tagged Offset Is: The offset within a Tagged Buffer on a Node. Suggest: The offset, in octets, from the start of a tagged buffer. Used with the tagged buffer model. Change: wording to make consistent with MO. Capitalization [hvs] Change not accepted. 28. Editorial, low, 4.3 under ULP Buffer Is: A buffer owned above the DDP Layer and advertised to the DDP Layer either as a Tagged Buffer or an Untagged ULP Buffer. Suggest: A buffer owned above the DDP layer and advertised to the DDP layer either as a tagged buffer or an untagged ULP buffer. Change: Capitalization [hvs] Change not accepted. 29. Editorial, low, 4.3 under ULP Message Length Is: is the total length of the ULP Payload contained in a DDP Message. Suggest: The total length, in octets, of the ULP payload contained in a DDP message. Change: removed leading is, added octets, changed capitalization. [hvs] Change accepted. 30. Editorial, low, 4.3 under Untagged Buffer Is: A buffer that is not explicitly Advertised to the Remote Peer. Suggest: A buffer that is not explicitly advertised to the remote peer. Change: capitalization [hvs] Change not accepted. Defined term are used with capitalization. 31. Editorial, low, 4.3 under Untagged Buffer Model Is: A DDP data transfer model used to transfer Untagged Buffers from the Local Peer to the Remote Peer. Suggest: A DDP data transfer model that preserves ULP message boundaries from the remote peer while keeping the local peers buffers anonymous. (See clause 3.1) [hvs] Change not accepted. 32. Editorial, low, 4.3 under Untagged DDP Message Is: A DDP Message that targets an Untagged Buffer. Suggest: A DDP message containing a queue number. Messages which contain queue numbers use the untagged buffer model. Change: reworded to remove the word "targets" [hvs] Change not accepted. The suggested text does not make the definition better. Barry Reinhold Lamprey Networks [email protected] (603) 868-8411