Randy's comments on the DLUR and HPR MIBs
"R.E. (Robert) Moore (254-4436)" <REMOORE@ralvm6.vnet.ibm.com> Tue, 11 March 1997 16:27 UTC
Received: from cnri by ietf.org id aa06587; 11 Mar 97 11:27 EST
Received: from brittany.cisco.com by CNRI.Reston.VA.US id aa11171; 11 Mar 97 11:27 EST
Received: from hubbub.cisco.com (hubbub.cisco.com [198.92.30.31]) by brittany.cisco.com (8.6.12/8.6.5) with ESMTP id IAA09163 for <extdom.snanaumib@aliashost.cisco.com>; Tue, 11 Mar 1997 08:21:49 -0800
Received: from VNET.IBM.COM ([199.171.26.4]) by hubbub.cisco.com (8.8.4-Cisco.1/CISCO.GATE.1.1) with SMTP id IAA09673 for <snanaumib@external.cisco.com>; Tue, 11 Mar 1997 08:21:37 -0800 (PST)
Message-Id: <199703111621.IAA09673@hubbub.cisco.com>
Received: from RALVM6 by VNET.IBM.COM (IBM VM SMTP V2R3) with BSMTP id 5946; Tue, 11 Mar 97 11:21:08 EST
Date: Tue, 11 Mar 1997 10:59:18 -0500
From: "R.E. (Robert) Moore (254-4436)" <REMOORE@ralvm6.vnet.ibm.com>
To: aiw-appn-mibs@raleigh.ibm.com, snanaumib@external.cisco.com
Subject: Randy's comments on the DLUR and HPR MIBs
As I promised, here are Randy's comments on our two MIBs. We're going to discuss these at AIW 13, but as I said earlier, we should try to start any significant discussions beforehand on the mailing lists. Bob Moore ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Date: Tue, 25 Feb 1997 11:17:11 -0800 From: Randy Presuhn <rpresuhn@peer.com> Message-Id: <199702251917.AA203998231@dorothy.peer.com> To: kostick@qsun.att.com Subject: Review of SNANAU DLUR MIB Cc: clouston@cisco.com, mo@uunet.uu.net, remoore@ralvm6.vnet.ibm.com, sob@harvard.edu Mime-Version: 1.0 Content-Type: text/plain; charset=X-roman8 Content-Transfer-Encoding: 7bit Hi Deirdre - As requested, I have reviewed <draft-ietf-snanau-dlurmib-00.txt> I think it needs another round, with special attention to the indexing issues and making its relationship to the other SNA MIBs more explicit. My comments are attached. Item: 0 Severity: minor Page: general Clause: general, especially 4 Rationale: Scattered through the MIB are references to other SNA- related MIBs, including SNANAU and APPN. Clearly these fit together in some grand scheme. Proposal: Add text to clause 4 (perhaps a new clause 4.2) outlining the relationship of this MIB to other MIBs. In particular, it would be helpful to describe any "referential integrity" constraints ("shared" indexes, "foreign keys", and relationships (dependencies?)) between entries in these other MIBs and this MIB. Item: 1 Severity: editorial Page: 2 Clause: 4 Rationale: Missing definition of acronym. Proposal: In the third paragraph, define "PU" Item: 2 Severity: editorial Page: 2 Clause: 4 Rationale: Missing definition of abbreviation. Proposal: It would be helpful to spell out what CPSVRMGR stands for. (Control Point Server Manager?) Item: 3 Severity: minor Page: 3 Clause: 4 Rationale: Completeness. Proposal: Identify which conditions causing APPN MIB traps affect DLUR resources. In particular, how would one go from the varbinds in an APPN MIB trap to a specific table row in the DLUR MIB? Item: 4 Severity: editorial Page: 3 Clause: 4.1 Rationale: Clarity. The word "group" appears to be used here and in the following clause (4.1.1) in both the English and the RFC 1904 senses. Proposal: Replace "groups" with "collections" in the first sentence. Delete "groups" in the last sentence. Item: 5 Severity: minor Page: 6 Clause: 5 Rationale: Some (broken) MIB compilers choke on lines of dashes. Proposal: Fix the line immediately following "Textual Convention". Item: 6 Severity: question/major Page: 6 Clause: 5 Question: Is it intended that there shall never be more than one APPN node per managed system? Is it impossible for a single managed system to be multiple APPN nodes? If not, then dlurNodeCpName through dlurDefaultDefPrimDlusName would need to be in a table. Item: 7 Severity: Major Page: 8-9 Clause: 5 Rationale: Consistency and accuracy. The DESCRIPTION for dlurDefaultDefBackupDlusTable describes the entries as being ordered by preference. The DESCRIPTION for dlurDefaultDefBackupDlusEntry says the index is arbitrary. The SYNTAX for dlurDefaultDefBackupDlusIndex limits the index range without explanation, and its DESCRIPTION does not give any clue that this index has preference semantics. Proposal: Modify the DESCRIPTIONs for dlurDefaultDefBackupDlusEntry and for dlurDefaultDefBackupDlusIndex to reflect the actual indexing semantics. Explain the range constraints on dlurDefaultDefBackupDlusIndex. Item: 8 Severity: editorial Page: 9 Clause: 5 Rationale: It is not clear what is meant by the three uses of the phrase "DLUR-unique" on this page. Proposal: Clarify. :-) Item: 9 Severity: editorial Page: 10 Clause: 5 Rationale: Missing reference (SNANAU MIB) in dlurPuName DESCRIPTION. Proposal: Add the reference to clause 7. Item: 10 Severity: MAJOR Page: 10 Clause: 5 Rationale: Imaginary object? I am unable to find a definition for snaNodeSscpSuppliedName (referenced in the description of dlurPuSscpSuppliedName) in RFC1666 or in <draft-ietf-snanau-naumib-00.txt> (which has expired, by the way). Proposal: Correct the reference, adding to the references section as needed. Item: 11 Severity: Minor Page: 11 Clause: 5 Rationale: The meanings of the possible values in this enumeration for dlurPuStatus are missing. Proposal: Explain the significance of each of the possible values in the DESCRIPTION of dlurPuStatus. Item: 12 Severity: Question Page: 12 Clause: 5 Rationale: Verification of understanding. Is this how the states of the contention winner and contention loser are combined for dlurPuDlusSessnStatus? Is the loss of information acceptable? | pA | pI | A | X = !(pA | pI | A) ---+----+----+---+---------------- pA | 2 | 2 | 2 | 2 ---+----+----+---+---------------- pI | 2 | 4 | 4 | 4 ---+----+----+---+---------------- A | 2 | 4 | 3 | 1 ---+----+----+---+---------------- X | 2 | 4 | 1 | 1 ---+----+----+---+---------------- Item: 13 Severity: minor Page: 12 Clause: 5 Rationale: The phrase "if present" in the DESCRIPTION of dlurPuActiveDlusName is potentially ambiguous. Proposal: Replace "If present" with "If the length is not zero". Item: 14 Severity: Major Page: 14 Clause: 5 Rationale: See item 7 above. The DESCRIPTIONs for the indexing of dlurPuDefDackupDlusTable, dlurPuDefBackupDlusEntry, and dlurPuDefBackupDlusIndex are inconsistent. Proposal: Modify these DESCRIPTIONs to describe the intended preference semantics and the motivation for the range constraints. Item: 15 Severity: Question Page: 15 Clause: 5 Question: The definition of dlurDlusName would indicate that there is an architectural limit of one CPSVRMGR pipe between any DLUS/DLUR pair. If this is not correct, the indexing structure of this table would need to be changed. Item: 16 Severity: Question Page: 16 Clause: 5 Question: For the dlurDlusSessnStatus, is the state combination matrix like that for item 12 above, and is the information loss acceptible? Item: 17 Severity: Major Page: 19 Clause: 8 Rationale: Vacuous security considerations sections are no longer acceptible. Proposal: Determine the risks of exposing the information provided by this mib are, so that appropriate SNMP protections can be configured. Note that there are no writable or creatable objects. --------------------------------------------------------------------- Randy Presuhn BMC Software, Inc. (Silicon Valley Division) Voice: +1 408 556-0720 (Formerly PEER Networks) http://www.bmc.com Fax: +1 408 556-0735 1190 Saratoga Avenue, Suite 130 Email: rpresuhn@bmc.com San Jose, California 95129-3433 USA --------------------------------------------------------------------- In accordance with the BMC Communications Systems Use and Security Policy memo dated December 10, 1996, page 2, item (g) (the first of two), I explicitly state that although my affiliation with BMC may be apparent, implied, or provided, my opinions are not necessarily those of BMC Software and that all external representations on behalf of BMC must first be cleared with a member of "the top management team." --------------------------------------------------------------------- Date: Fri, 28 Feb 1997 20:09:10 -0800 From: Randy Presuhn <rpresuhn@peer.com> Message-Id: <199703010409.AA280199350@dorothy.peer.com> To: kostick@qsun.att.com Subject: SNANAU HPR MIB review Cc: clouston@cisco.com, jhalpern@newbridge.com, mo@uunet.uu.net, remoore@ralvm6.vnet.ibm.com, sob@harvard.edu Mime-Version: 1.0 Content-Type: text/plain; charset=X-roman8 Content-Transfer-Encoding: 7bit Hi Deirdre - I've completed my initial review of <draft-ietf-snanau-hprmib-00.txt> My comments are attached. I think it needs another round of comment. My biggest concerns are: - relationship of indexing structure to APPN MIB - missing security considerations - internal inconsistency on support for switching to a specified path. - interactions resulting from ability to turn counting off and on. Item: 1 Severity: editorial Page: 2 Clause: 4 Proposal: Delete "the proposed set of" in the first sentence of the first paragraph in this clause. Item: 2 Severity: question Page: 3 Clause: 4 Question: It the sentence at the very top of the page, is the path really always going to be optimal? Item: 3 Severity: editorial Page: 3 Clause: 4 Rationale: Complete references are helpful. Proposal: Provide a reference for the ARB algorithm. Item: 4 Severity: editorial Page: 3 Clause: 4 Rationale: Organization. Proposal: Move the paragraph beginning "The HPR terms and..." to the beginning of the section. It would be valuable to provide sources for the HPR architecture beyond an FTP site. Item: 5 Severity: editorial Page: 3 Clause: 4 Rationale: Terms should be defined. Proposal: Define "Transport Tower". (third bullet on page) Item: 6 Severity: minor Page: 3 Clause: 4 Rationale: In the last bullet in the second group of bullets, there is reference to an APPN MIB trap. It would be helpful to explain how one would get to this MIB from the information carried by that trap. Proposal: Add text explaining relationship between APPN MIB traps and this MIB. Item: 7 Severity: minor Page: 3 Clause: 4 Rationale: The relationships between this MIB and the APPN MIB should be explained more fully. At the very least, the indexing relationships between the MIBs should be spelled out, particularly if there are any potential referential integrity issues when objects are used as foreign keys into tables in other MIBs. Proposal: Identify the objects in this MIB used as keys into the APPN MIB. Identify the objects from the APPN MIB used as keys into this MIB. Item: 8 Severity: editorial Page: 3 Clause: 4.1 Rationale: The term "group" is used in both the RFC 1904 sense and the English sense. Proposal: Replace "groups" with "collections" in the first sentence. Item: 9 Severity: editorial Page: 4 Clause: 4.1.1 Rationale: Expand abbreviations - CP is used before being defined. Proposal: Add the expansion of CP to the first sentence. (It is defined later at the bottom of the page.) Item: 10 Severity: editorial Page: 4 Clause: 4.1.2 Rationale: A numbered list with only one item isn't particularly helpful. Proposal: Replace "1) hprAnrRoutingTable ... This table" with "The hprAnrRoutingTable" Item: 11 Severity: editorial Page: 4 Clause: 4.1.3 Rationale: A numbered list with only one item isn't particularly helpful. Proposal: Replace "1) hprNceTable ... This table" with "The hprNceTable" Item: 12 Severity: editorial Page: 5 Clause: 4.1.4 Rationale: Expand abbreviations Proposal: Explain what TCID is in the third paragraph. Item: 13 Severity: editorial Page: 6 Clause: 5 Rationale: readability Proposal: add a blank line after the last of the IMPORTS Item: 14 Severity: minor Page: 7 Clause: 5 Rationale: Some MIB compilers choke on lines of dashes Proposal: Get rid of the line of dashes Item: 15 Severity: editorial Page: 7 Clause: 5 Rationale: Clarity. Proposal: Insert the words "correspondingly indexed" before "hprRtpCounterDisconTime". Item: 16 Severity: question Page: 7 Clause: 5 Question: It is not clear what is meant by "corresponds to" in the DESCRIPTION of hprNodeCpName. If the value is the same, why is this object needed? If it can be different, how are they related? What does this object DO? Item: 17 Severity: minor Page: 7 Clause: 5 Rationale: The hprUpTime object is not used. Proposal: Use it or delete it. There are several references in this MIB to appnNodeCounterDisconTime that should probably be references to this object. Item: 18 Severity: major Page: 8 Clause: 5 Rationale: The description of switchToPathSupported is in direct conflict with the text on page 19 for hprRtpPathSwitchTrigger, which in the last paragraph of its DESCRIPTION says this module provides no such feature. Proposal: Either delete this value or add the necessary support object. Item: 19 Severity: minor Page: 8 Clause: 5 Rationale: It is unclear why hprAnrsAssigned uses an object from the appn MIB as its discontinuity indicator. Proposal: Explain it or use the hprUpTime object instead. Item: 20 Severity: minor Page: 8 Clause: 5 Rationale: missing UNITS. Proposal: Add UNITS to hprAnrsAssigned. Item: 21 Severity: minor Page: 9 Clause: 5 Rationale: No rationale is given for counters being set to / held at zero when hprAnrCounterState is notActive. Proposal: Explain the reason for this additional constraint or remove it. Item: 22 Severity: minor Page: 9 Clause: 5 Rationale: The start-up value for hprAnrCounterStateTime should be defined. Proposal: Add text describing initial value for this object. Item: 23 Severity: editorial Page: 9 Clause: 5 Rationale: It is not clear what is meant by "TG information objects". Proposal: Identify which objects are meant by "TG information objects" Item: 24 Severity: minor Page: 11 Clause: 5 Rationale: It looks like there may be a slight ambiguity to the use of the value 0 with hprAnrOutTgNum, since this is a legitimate value for appnLocalTgNum in the APPn MIB. Proposal: Some explanatory text might be helpful here, particularly if I've been led astray. :-) Item: 25 Severity: editorial Page: 11 Clause: 5 Rationale: In the DESCRIPTION of hprAnrPacketsReceived, the word "total" may be misleading, unless there is a strong assumption that the counter is zero-based. Also, it should be explicit that the hprAnrCounterDisconTime to monitor is the one in the same row. Proposal: Replace "total" with "count"; add "in the same row" at the end of the description. Item: 26 Severity: minor Page: 11 Clause: 5 Rationale: need UNITS Proposal: add UNITS to hprAnrPacketsReceived Item: 27 Severity: minor Page: 12 Clause: 5 Rationale: The definition of hprAnrCounterState's notActive value on page 9 means that these counters experience discontinuity not only when they are turned on, but also when they are turned off. Proposal: Add "or off" to the end of the description for hprAnrCounterDisconTime. Item: 28 Severity: minor Page: 12 Clause: 5 Rationale: In the description of hprAnrCounterDisconTime, the word "later" may be problematic. If sysUpTime is reset for some reason, recording its value here may result in a SMALLER value. In general, one can't make assumptions about this variable being non-decreasing. What sysUpTime reports is not really a time, but rather an interval. Proposal: Replace "later" with "most recent". Item: 29 Severity: minor Page: 13 Clause: 5 Rationale: The relationship of the hprNceId index to other tables and MIBs should be described, if there is any. Proposal: ? Item: 30 Severity: minor Page: 13 Clause: 5 Rationale: The enumeration for hprNceType is awkward and unenlightening. Proposal: Use the BITS construct instead. Item: 31 Severity: minor Page: 14 Clause: 5 Rationale: The enumeration for hprNceDefault is awkward and unenlightening. The "noDefault" value looks like a hack. Proposal: Use the BITS construct instead. If no bits are set, does this adequately represent the "noDefault" case? If so, then a TEXTUAL-CONVENTION for this and the previous object may be appropriate. Item: 32 Severity: editorial Page: 15 Clause: 5 Rationale: The DESCRIPTION of hpr NceInstanceId is not particularly enlightening. Proposal: Is there anything more to say about this thing? Can its values be used as foreign keys into other tables? Item: 33 Severity: minor Page: 15-25 Clause: 5 Rationale: The following items lack UNITS clauses: hprRtpGlobeConnSetups hprRtpLivenessTimeouts hprRtpShortReqTimeouts hprRtpMaxSendRate hprRtpMinSendRate hprRtpCurSendRate hprRtpSendPackets hprRtpRecvPackets hprRtpSendBytes hprRtpRecvBytes hprRtpRetrPackets hprRtpPacketsDiscarded hprRtpDetectGaps hprRtpRateReqSends hprRtpOkErrPathSws hprRtpBadErrPathSws hprRtpOkOpPathSws hprRtpBadOpPathSws Proposal: Add appropriate UNITS. Also, determine whether the phrase "total number" should be replaced by "count of" to avoid assumptions on zero-based counters. Item: 34 Severity: minor Page: 15 Clause: 5 Rationale: It is not clear why the discontinuity indicator for hprRtpGlobeConnSetups should be in the APPN MIB. Proposal: replace from "appnNodeCounterDisconTime" to the end of the sentence with "hprUpTime" Item: 35 Severity: minor Page: 16 Clause: 5 Rationale: For hprRtpGlobeCtrState, it is not clear why notActive should set/hold counters to zero. Proposal: Explain or delete "their values are always zero" Item: 36 Severity: minor Page: 16 Clause: 5 Rationale: The description for hprRtpGlobeCtrStateTime claims that this objects can be used to relate the time of a state change to the last time of APPN node re-initialization. It is not clear how this is possible, given the other objects in this MIB and the APPN MIB. Proposal: Explain it. :-) Item: 37 Severity: Question Page: 16 Clause: 5 Question: The comment preceding hprRtpTable and the index structure of this table implies an architectural limit of one HPR transport tower (whatever that is) per node. Is this deliberate and appropriate? Item: 38 Severity: editorial Page: 17 Clause: 5 Rationale: cosmetics. hprRtpRecvPackets, hprRtpSendPackets, hprRtpRecvBytes, and hprRtpSendBytes appear in a different order in the SEQUENCE and the following definitions. Proposal: Make the orders match. Item: 39 Severity: minor Page: 18 Clause: 5 Rationale: The relationship of the hprRtpLocNceID and hprRtpLocTcId indexes to objects in the APPN MIB should be spelled out in the descriptions. What is the relationship of these to the APPN appnLsTable, appnIsInTable, and appnIsRtpTable? Proposal: Describe the relationships. Item: 40 Severity: question Page: 18 Clause: 5 Question: For hprRtpRemCpName, is it possible for the remote node name to be unknown? Item: 41 Severity: question Page: 18 Clause: 5 Question: For hprRtpRemNceId, is it possible for the remote NCE to be unknown? Item: 42 Severity: question Page: 18 Clause: 5 Question: For hprRtpRemTcid, is it possible for the remote TCID to be unknown? Item: 43 Severity: minor Page: 19 Clause: 5 Rationale: The third paragraph of the DESCRIPTION of hprRtpPathSwitchTrigger suggested unusual behaviour. RFC 1905 clause 4.2.5(10) [page 20] would suggest that the SNMPv2 inconsistantValue (mapping to SNMPv1 badValue) would be a more appropriate response. Proposal: Align the behaviour with SNMP. This would be as simple as deleting the third paragraph. Item: 44 Severity: major Page: 19 Clause: 5 Rationale: The last paragraph of the DESCRIPTION of hrpRtpPathSwitchTrigger directly contradicts the text on page 8 for switchToPathSupported. Proposal: Add objects to support the desired capability. Item: 45 Severity: minor Page: 22 Clause: 5 Rationale: For the rates hprRtpMaxSendRate, hprRtpMinSendRate, and hprRtpCurSendRate, is the period over which the rates are computed implementation dependent? The dt overwhich rates are computed will have significant impact on both the high and low water marks, particularly immediately after connection setup. Is any smoothing applied? Proposal: If there is a standard interval over which rates are computed, provide it. If not, say that it's an implementation choice. Item: 46 Severity: minor Page: 23 Clause: 5 Rationale: Is the smoothing algorithm for hprRtpSmRdTripDelay implementation-dependent? Proposal: If so, say so. Otherwise, outline the characteristics the of smoothing algorithm to be used. Item: 47 Severity: minor Page: 26 Clause: 5 Rationale: In the description of hprRtpCounterDisconTime, the word "later" may be problematic. If sysUpTime is reset for some reason, recording its value here may result in a SMALLER value. In general, one can't make assumptions about this variable being non-decreasing. What sysUpTime reports is not really a time, but rather an interval. Proposal: Replace "later" with "most recent". Item: 48 Severity: minor Page: 26 Clause: 5 Rationale: Due to the description of hprRtpGlobeCtrState, the last sentence of hprRtpCounterDisconTime is not complete. Proposal: Add "or off" to the end of the DESCRIPTION. Item: 49 Severity: minor Page: 27 Clause: 5 Rationale: The relationship of the indexes for hprRtpStatusEntry to the other tables in this MIB and the APPN MIB should be described. Proposal: Add text describing relationship. Does the indexing structure of this table align with anticipated usage scenarios? Item: 50 Severity: minor Page: 28 Clause: 5 Rationale: The description of hprRtpStatusIndex needs adjustment. The intent appears to be to order history entries for a given NCEID/TCID pair chronologically, or possibly to provide a global chronological ordering of all history entries (though not supported by the index structure) Proposal: possible replacement text might look like: "Each entry in this table is assigned a unique value for this index, starting with zero and going through the integers consecutively." Wrap behaviour should also be covered. Item: 51 Severity: questions Page: 28 Clause: 5 Question: Will hprRtpStatusRemCpName always be known? Question: Will hprRtpStatusRemNceId always be known? Question: Will hprRtpStatusRemTcId always be known? Item: 52 Severity: editorial Page: 29 Clause: 5 Rationale: Clarity. Proposal: For hprRtpStatusNewRscv and hprRtpStatusOldRscv, explain what a 0-length octet string means. Item: 53 Severity: editorial Page: 33 Clause: 5 Rationale: consistency Proposal: re-order hprRtpRecvPackets, hprRtpSendPackets, hprRtpRecvBytes, hprRtpSendBytes to match the order of the definitions. Item: 54 Severity: editorial Page: 35 Clause: 7 Rationale: The url for [5] appears to go beyond the RFC margins. A "real" reference would be nice. Proposal: ? Item: 55 Severity: major Page: 36 Clause: 8 Rationale: Vacuous security considerations are a no-no. Proposal: Identify the security implications of the settable objects in this mib. Point to generic SNMP privacy/ authentication stuff. Identify any information in this MIB which may, through exposure, be a security problem. --------------------------------------------------------------------- Randy Presuhn BMC Software, Inc. (Silicon Valley Division) Voice: +1 408 556-0720 (Formerly PEER Networks) http://www.bmc.com Fax: +1 408 556-0735 1190 Saratoga Avenue, Suite 130 Email: rpresuhn@bmc.com San Jose, California 95129-3433 USA --------------------------------------------------------------------- In accordance with the BMC Communications Systems Use and Security Policy memo dated December 10, 1996, page 2, item (g) (the first of two), I explicitly state that although my affiliation with BMC may be apparent, implied, or provided, my opinions are not necessarily those of BMC Software and that all external representations on behalf of BMC must first be cleared with a member of "the top management team." ---------------------------------------------------------------------
- Randy's comments on the DLUR and HPR MIBs R.E. (Robert) Moore (254-4436)