DefCore Misconceptions Part 2: Debunking Advisory Status

DefCore isn’t a new creature in the OpenStack community: it’s been in discussion since at least 2013. However it was only earlier in 2015 that adherence to DefCore Guidelines became a requirement for products that want to use the OpenStack name and OpenStack Powered logo. As a result, a lot more people are now interested in DefCore than in the past. As we’ve started receiving a lot more feedback about DefCore, it’s become apparent that there are a few misconceptions out there from folks who are new to DefCore. Personally I held some of those same misconceptions when I started participating in DefCore about 9 months ago. This series of posts is intended to lay some of those misconceptions to rest and help folks learn more about DefCore is, how it works, and what it’s trying to accomplish.

In DefCore parlance, a Guideline is essentially a document that lists what capabilities an OpenStack Powered(TM) product must expose to end users. For each one of those Capabilities, there is also a list of tests that the product (which could be a public cloud, a distribution, an appliance, or a managed offering) must pass in order to prove that it supports those Capabilities. Individual Capabilities can be marked as being in one of four states: required, advisory, deprecated, or removed. As soon as you read those four words, a few ideas about the lifecycle of Capabilities probably starts to materialize in your head. Today I’ll talk about one of those states (we’ll save the others for a later post): advisory.

A Capability starts off life when the DefCore Committee considers adding and removing Capabilities to it’s Guidelines, which happens once every six months and is offset by a couple of months from OpenStack software releases. This part of the lifecycle is pretty subjective: generally speaking, in order to be proposed someone has to think that there’s a strong enough argument for including a capability to bring it up for discussion. Typically the list of candidates has been created by having members of the DefCore Committee chat with PTL’s or other members of the technical contributor community while also injecting their own opinions, though anyone can submit ideas. You don’t have to be sure the candidate is a slam-dunk for inclusion, you just need to think it’s at least worth talking about with a wider audience.

The next step is that the candidate Capabilities are proposed for scoring. In order to decide what Capabilities make it into Guidelines, the DefCore Committee has a set of 12 Criteria that it measures Capabilities against (mostly are trailing indicators of market acceptance). In the early days of the Committee, scoring was done via live audio meetings with a Google spreadsheet and was, IMHO, a somewhat inexact process (which is fine: bootstrapping a new thing is always challenging and your processes should naturally improve and mature over time as long as you incorporate feedback and lessons learned). In the current cycle, the DefCore Committee modified it’s process for scoring so that scoring is now done in Gerrit. This basically involves filling out a simple worksheet for the Capabilities you want to propose, and using a simple script to tally up a total score (it also emits a CSV version if you’re more accustomed to using a spreadsheets for this sort of thing). If a Capability scores high enough, you can include modifications to the still-under-construction next Guideline in the same patch. This allows everyone to see what candidates are being proposed, some initial thinking behind how they score to get the conversation started, and exactly what tests are being proposed for those candidates. You can have a look at several proposals for new capabilties that are currently going through this process.

You may notice that for each Capability added to a Guideline, there’s a field for status. Very importantly, when a new Capability is added to a Guideline for the first time, it must be listed in advisory state rather than required state according to DefCore’s Board-approved operating rules. So, what exactly does advisory mean? Does it mean that the Capability in question is going to be required in the future and that vendors should get busy adding it to their product offerings if they don’t include it today? Not quite.

There is actually a definition of advisory status included in the DefCore lexicon, and it’s fairly simple: it means that a Capability has been suggested for the next Guideline. Note that word: suggested. Essentially, the DefCore Committee is saying to the world: “hey ladies and gents, we’re thinking about requiring this Capability in the future–what do you think about that?” In other words:

“Advisory” status means that the DefCore Committee thinks there’s good reason for requiring a Capability and would like to solicit feedback from the community about whether that’s actually the case.

So, just because a Capability is marked as advisory in one Guideline does not mean it will automatically graduate to required status in the next Guideline. In many cases that might actually happen if the scoring was done properly and there’s not much argument about a given Capability (and in such cases the advisory state also serves as a handy heads-up to vendors to check whether they’re supporting that Capability). But the Committee might also get feedback such that the Capability actually isn’t a good choice. That might be for any number of reasons: perhaps the Capability is actually specific to a particular backend driver, maybe it’s not actually as widely deployed as was assumed, or perhaps it’s not as widely used by clients or tools as was thought, etc etc.

Getting the scoring right for a given Capability is actually a fairly hard and labor-intensive process that involves doing a lot of research, and sometimes good information isn’t readily available. The advisory status actually serves as means of stepping up the conversation so that more information can be gathered and more opinions can be heard, and it provides a minimum of six full months for that conversation to happen. Notice the word minimum there. If the community is divided, it’s entirely possible the Capability might stay in advisory status while more discussion is had or while other adjustments (to code, to products, to tests, to clients, etc) are made. Alternately, the Capability might simply be dropped in the next Guideline after discussion indicates it’s really not a good candidate. Or, it may graduate to required status if there aren’t good grounds for dropping it or prolonging the advisory period.

It’s worth mentioning that one clue that advisory Capabilities don’t always become required is that the two status have different bars to meet when scoring a Capability. The DefCore Committee picks a cut-off score that Capabilities must meet in order to become required. That score can change from cycle to cycle as weights to Criteria are also adjusted, but in the past the cut-off score has hovered around 74. Though not strictly required to do so, the DefCore Committee has also generally applied a minimum score for advisory Capabilities: if a candidate can’t muster at least 50 points, there’s probably not much point in having a larger discussion about it.

I should also note that there are all kinds of “pressure relief valves” built into the DefCore process, of which the six-month advisory period is just one. For example, if you read the timeline section of DefCore’s operating procedures that I mentioned earlier, you’ll notice that Capabilities are generally selected for advisory status by the time a Summit rolls around (with an implicit assumption that Summits are a good opportunity to talk about what’s being proposed so everyone is aware of it), but the proposal doesn’t go to the Board of Directors for a full three months afterward. That means there’s actually several months even before a Capability becomes required during which it’s merits can be discussed (not to mention the discussion that happens when the initial scoring patch is submitted). And once a Capability does become required, it can be flagged if there’s sufficient reason to think it shouldn’t be required.

So there you have it: advisory status doesn’t mean that a Capability will necessarily be required in the future, just that there is a chance it will be subject to discussion by all segments of the community.

Questions? Comments? Drop by #openstack-defcore on IRC or drop the DefCore Committee a line at defcore-committee@lists.openstack.org.