Keith Moore wrote: >>The only way to releive work is to distribute it, not concentrate it. > > False. You can also relieve work while keeping throughput constant by > reducing overhead. > > Distributing work often reduces throughput by creating more overhead. > Only a few applications are embarassingly parallel - most do not scale > well as the number of PEs is increased. Agreed, but this isn't a computer; it runs very much like a program committee, for which there are hundreds (if not thousands) of examples of how to organize things more effectively. >>ADs often don't have the time to track all the WGs and can end up >>holding docs unnecessarily as a result. And they're not uniquely privvy >>to the concept of compromise or negotiation; their "power" should come >>from technical expertise, which is distributed in the IETF, not from >>fiat (I didn't see white smoke from the Nomcom). > > Just by doing their jobs, the ADs are inherently exposed to a wider > range of issues than other IETF participants, probably including those > in the IAB. Nobody else gets to concurrently see the problems that > exist at every level of the protocol stack. Just by doing their jobs ADs focus on the area they direct; they can be(and have been) just as myopic to that area as individuals working on problems. Very few of us mortals do NOT see the problems at other levels of the protocol stack. >>ADs are not the only ones who can look or work at mutliple layers; in >>fact, their focus works against them in many cases in this regard. > > That's not "fact", that's just your conjecture. And my experience > contradicts it. It's my experience that contradicts yours, not a conjecture. I wouldn't bother with any of my observations were they hypothetical. > ADs tend to become less focused over time, while these > days many individuals in IETF work only in a particular niche. Every > AD sees considerably more documents from outside his area than inside. > They don't review every document of course, but they do participate in > discussions about every document. So you're arguing for a single monolithic PC where only PC members provide reviews (common among workshops), whereas I am arguing for focused groups, which is the more common program committee model for conferences, esp. where scale is an issue. >> And there are PLENTY of docs that need review that aren't standards >> track that the IESG is just plain holding up. > > Generally, those docs need review because they will be viewed as > standards or as conflicting with standards. And not surprisingly, > working group and non-standards track documents have priority. Those docs are reviewed to see IF they have such issues; it's not possible to know whether such conflicts MIGHT exist until the doc is reviewed. > Something that people seem to miss is that for any document that goes > before IESG, only one AD has to review that document. The other ADs > are free to say "no objection". Those who read the document do so > because they believe that the document needs scrutiny, not because > their jobs require them to do so. IESG people are more acutely aware > of the potential for documents to create problems than most other IETF > participants, so they're going to take time to review documents that > seem to have the potential to create problems. The assertion that the IESG is more acutely aware than plenty of people in the IETF is a nice assertion, but that's all it is. There's plenty of evidence that they aren't aware of interactions or prior work of (in some cases their own) WGs. > To the extent that ADs can entrust those reviews to other people, they > are able to do so and always have been able to do so. And this works > well for small documents that only affect a particular niche. But for > many documents there aren't many reviewers who have both the time/ > energy and a sufficiently broad perspective to replace the AD's own > review and to make the IESG's deadlines. Some of the documents I > reviewed while on IESG took ten to twenty hours of intensive time to > review - they were that long and complex and it required multiple > passes to see whether the loose ends were tied up. IESG people get > committment from their employers to support that work. Try saying to > someone whose employer hasn't made that committment - not a student, > but someone who is experienced and has a broad perspective, "I need you > to take ten or twenty hours away from your day job to review this > document, and I need your review by a week from this Thursday". How > often do you think that will work? About 5,000 times a year for Infocom; plenty of us spend time on program committees and provide plenty of reviews that take many hours in *some* cases. >>If you can find a dozen or so people to stop >>what they're doing and be ADs, you can certainly find more people who >>will participate if they can do so at a lower level of commitment. > > No, it doesn't follow. It follows as much as other assertions that are being tossed about as fact: reducing the level of effort increases the number of people who are willing to volunteer that effort. >>>The Apps area had a directorate for many years with the >>>idea that some reviews could be offloaded to the directorate. It didn't >>>work very well, because those people were busy too, and it was hard to >>>find people who could reliably turn around a review of a difficult >>>document in time for the IESG telechat. They could (and did!) review the >>>short/easy/noncontroversial documents, but there aren't that many of >>>them, and they're not the ones that take up all of an AD's time. >> >>OK, so we have ONE example where it didn't work. Let's throw it out. ;-) > > sure, let's ignore the evidence and depend on conjecture :) The evidence - overwhelming - is that this process works in plenty of other places. The personal experience is that a single case has been shown not to work. We have cases where Nomcom failed, ADs failed, etc. - are we ignoring the 'evidence' to keep those processes? >>The issue with "directorates" is that they turn into papal conclaves (to >>reuse the allusion). They need to be more open, but they can work fine. > > Directorates can be structured however the ADs want them to be > structured. We tried a variety of structures in Apps, including > completely open groups focused in particular areas (messaging, > directories, web). They were even less useful than the "papal > conclave" style directorates that they replaced. Well the current system of ADs isn't useful in some cases either; something has to change somewhere, and no system is going to be perfect. The question is whether it spread the load and was *sufficient*, not whether it was more useful than overloading a single AD. As noted before, single ADs have other properties - some have been in the same role for a decade - that give them context that isn't as much a function of being an AD as spending 80% of their career for the past 10 years reviewing IETF docs. >>>(Aside: I really do think that a limit on the number of documents, >>>and a page limit on normative technical specifications, might go a >>>long way toward lessening AD workload and also helping WGs produce >>>well-written specifications. But there would need to be some way of >>>making exceptions, as there will be a few documents that really do >>>need to be long.) >> >>Page limits are fine, but won't lessen the work, just the chunksize. > > which is why you need document limits also. Sure - so who reads the document proposals and decides which ones we review? Seems like a circular argument... >>While I can't speak for all WGs, I can speak for the ones I participate >>in more heavily, and they certainly never operate that myopically. What >>DOES happen is that we solicit cross-area review, or send authors to >>other WGs to present docs before the IESG review. > > If all WGs did frequent cross-area review, and documented that > review, IESG's job would be much easier. If they believe the documentation. But when IDs get sent to the IESG *with* documented cross-area review and get asked "did you ever present this to WG X?" - when WG X is in the list - things get just plain ridiculous. (and no, this isn't conjecture either) > In my experience, presenting > docs before other WGs doesn't work as effectively. It requires a > continuous dialogue. A problem with both approaches is that there's not > always another WG to represent the area of concern - which is an > artifact of our "everything has to be a working group" mentality. So now you're arguing that the ADs are there for content review, rather than to direct things to other WGs or groups. This is the problem I have with the current structure - it basically says "rough consensus, running code, AND the individual approval of the IESG". We don't believe in kings, and IMO, the IESG have too much king-like power in the current structure. >>Then the IESG says "did you take this to X?" - when we had years >>earlier. > > years earlier isn't good enough. the Internet changes too frequently. In the case I'm citing it was "we had years earlier, and had been working with ever since". >>The other issue is more of "hey, *I* have a technical problem with this >>doc", regardless of whether it has already had cross-area review and >>consensus. In which case we end up trying to appease the AD to sate >>their personal perspective, not the IETF or the Internet as a whole, and >>often at the expense (not to the credit) of the broad vision of the >>global Internet. > > I've seen two kinds of cases - one where the AD has a legitimate > technical problem with the document, often a problem that isn't visible > at the WG level because those people don't have a broad perspective > (or because the WG has actively engaged in denial about that problem). This is why ADs should participate in the WGs; I thought in general that they did (or at least tried to). THAT is where the cross-area stuff should occur, not later when the WG has finished (and sometimes disbanded). > I've also seen cases where an AD objected to a document on what > appeared to be a mere whim, without stating technical reasons for > doing so, or without brooking any kind of technical discussion. These > days where the ADs comments are visible to anyone who wants to > look at the document tracker Not all of the detailed comments. Some comments refer to private discussions. >, I suspect it's harder for an AD to get > away with that - and there are more grounds for appeal when the AD > does try that. But because there are often cases where an AD sees > things that the narrowly-focused WG simply cannot see, I don't think > we can dispense with AD review. And offhand I don't see how to > build up a pool of reviewers who have that same level of perspective. > That doesn't mean we shouldn't try, but it probably does mean that > we shouldn't depend on it. > > Keith As I don't want to depend on an equally myopic pool of ADs either. I wish I trusted the ADs perspective as much as you, but I simply don't; there are too many preconditions on AD selection that have nothing to do with expertise (availability, notably). Joe
Attachment:
signature.asc
Description: OpenPGP digital signature
_______________________________________________ Ietf@xxxxxxxx https://www1.ietf.org/mailman/listinfo/ietf