Re: server side describe
It seems to me that we need to get better at making decisions for things like that. If we keep on arguing for small things, it will simply be time consuming and painfull for everybody. In this case, the situation seems simple. Part of the group do not agree with the proposal. We just have to accept it and move on. On Fri, Apr 3, 2020 at 2:48 AM Joshua McKenzie wrote: > One thing probably worth thinking about: we're a mostly irascible lot to > begin with and there's a global pandemic and Human Race Lockdown. I don't > know about the rest of you but I'm starting from a pretty not-chill place > these days; trying to be mindful of that. > > So for this: if we require a protocol change there's a clear forcing > function to get it into 4.0 since a) it's minor and b) the time horizon for > the next major is very uncertain. > > If we go a route that doesn't require a protocol change it can just go into > the next patch release after 4.0 right? Is there an urgency to get it into > 4.0.0 I'm missing in this scenario? > > > On Thu, Apr 2, 2020 at 5:25 PM sankalp kohli > wrote: > > > Whether a feature is fully done and whether it validates or invalidate > > testing is not the point here. The point is that it is a feature and > > violates feature freeze. If someone brings in a feature which is almost > > done and does not invalidate testing then will we merge all of them to > 4.0? > > Lot of features can be merged then based on this justification!! > > > > > > Considering this is a small feature, I wont -1 on it. I will disagree > and > > commit. > > > > > > > > On Thu, Apr 2, 2020 at 1:04 PM Jon Haddad wrote: > > > > > Chris's original patch used a virtual table which didn't even require a > > > protocol change. To me, the difference between having a CQL describe > vs > > a > > > virtual table is unimportant, since it's only drivers that need to care > > > about it. I'm completely fine with the simpler implementation of a > > virtual > > > table. > > > > > > Quite a bit of Chris's patch also fixes our broken server side CQL > > > generation, something that affects correctness in our snapshots. So > > either > > > way most of the code needs to go in before we release 4.0. Adding a > > single > > > file that creates a new virtual table is so trivial I'm having a hard > > time > > > understanding the opposition. > > > > > > On Thu, Apr 2, 2020 at 12:56 PM Nate McCall > wrote: > > > > > > > So summarizing the salient points here: > > > > - client authors have worked around this mostly, but this would avoid > > > some > > > > duplication of effort for new features > > > > - this issues was tagged last year as being pertinent to 4.0 in > several > > > > threads about what was in scope > > > > - there is some development efforts required to review/merge/update > > these > > > > patches and we are trying to release > > > > - the change is unintrusive > > > > - this is a change to the protocol > > > > > > > > Not having this doesnt effect me for $dayJob, but I want to reiterate > > > that > > > > it's a silly thing to leave to clients. Given we've previously scoped > > it > > > to > > > > 4.0, im still +1 on adding it. > > > > > > > > It's ok to have differing opinions. I'd like to see us disagree and > > > commit > > > > to a course of action either way as opposed to just letting it sit > more > > > > because we can't sort it out. > > > > > > > > > >
Re: server side describe
Someone once said: "I heard the expression recently that “there are ten ways to do this, and eight of them will work.” I think that applies to most of the code we write. We don't need to spend a lot of time discussing which of the eight is best; let’s trust the judgement of the original author and move forward. " Had we applied that principle to this JIRA the first time the patch was available in OCTOBER 2018, we wouldn't be having a conversation about whether or not it violates a freeze. On Wed, Apr 1, 2020 at 11:37 AM Jonathan Ellis wrote: > I think we should get serious about the so-called freeze. > > On Wed, Apr 1, 2020 at 1:27 PM Jon Haddad wrote: > > > Hey folks, > > > > I was looking through our open JIRAs and realized we hadn't merged in > > server side describe calls yet. The ticket died off a ways ago, and I > > pinged Chris about it yesterday. He's got a lot of his plate and won't > be > > able to work on it anytime soon. I still think we should include this in > > 4.0. > > > > From a technical standpoint, It doesn't say much on the ticket after > Robert > > tossed an alternative patch out there. I don't mind reviewing and > merging > > either of them, it sounded like both are pretty close to done and I think > > from the perspective of updating drivers for 4.0 this will save quite a > bit > > of time since driver maintainers won't have to add new CQL generation for > > the various new options that have recently appeared. > > > > Questions: > > > > * Does anyone have an objection to getting this into 4.0? The patches > > aren't too huge, I think they're low risk, and also fairly high reward. > > * I don't have an opinion (yet) on Robert's patch vs Chris's, with regard > > to which is preferable. > > * Since soon after Robert put up his PR he hasn't been around, at least > as > > far as I've seen. How have we dealt with abandoned patches before? If > > we're going to add this in the patch will need some cleanup. Is it > > reasonable to continue someone else's work when they've disappeared? > > > > Jon > > > > > -- > Jonathan Ellis > co-founder, http://www.datastax.com > @spyced >
Re: server side describe
This isn't a hill to die on or something to binding -1 for me personally. In a vacuum this merge is totally fine. The problem for me comes in if a merge like this is one of 10, or 50, or 100 things that are innocuous in isolation. IMO as long as we make sure this is the only cut we do to ourselves we won't face death by a thousand cuts on 4.0. In general I'm concerned about our rigor and discipline on restricting scope to get 4.0 out the door, but that's in no way unique to us as a project or 4.0 as a release; this has happened with every large software release I've ever worked on and always requires significant discomfort to lock down. It's unfortunate that the situation with this ticket stumbled across that sore point for me (and likely many of us), but I think we should keep that to the other thread about scope. On Fri, Apr 3, 2020 at 11:34 AM Jeff Jirsa wrote: > > Someone once said: > > "I heard the expression recently that “there are ten ways to do this, and > eight of them will work.” I think that applies to most of the code we > write. We don't need to spend a lot of time discussing which of the eight > is best; let’s trust the judgement of the original author and move > forward. " > > Had we applied that principle to this JIRA the first time the patch was > available in OCTOBER 2018, we wouldn't be having a conversation about > whether or not it violates a freeze. > > > On Wed, Apr 1, 2020 at 11:37 AM Jonathan Ellis wrote: > > > I think we should get serious about the so-called freeze. > > > > On Wed, Apr 1, 2020 at 1:27 PM Jon Haddad wrote: > > > > > Hey folks, > > > > > > I was looking through our open JIRAs and realized we hadn't merged in > > > server side describe calls yet. The ticket died off a ways ago, and I > > > pinged Chris about it yesterday. He's got a lot of his plate and won't > > be > > > able to work on it anytime soon. I still think we should include this in > > > 4.0. > > > > > > From a technical standpoint, It doesn't say much on the ticket after > > Robert > > > tossed an alternative patch out there. I don't mind reviewing and > > merging > > > either of them, it sounded like both are pretty close to done and I think > > > from the perspective of updating drivers for 4.0 this will save quite a > > bit > > > of time since driver maintainers won't have to add new CQL generation for > > > the various new options that have recently appeared. > > > > > > Questions: > > > > > > * Does anyone have an objection to getting this into 4.0? The patches > > > aren't too huge, I think they're low risk, and also fairly high reward. > > > * I don't have an opinion (yet) on Robert's patch vs Chris's, with regard > > > to which is preferable. > > > * Since soon after Robert put up his PR he hasn't been around, at least > > as > > > far as I've seen. How have we dealt with abandoned patches before? If > > > we're going to add this in the patch will need some cleanup. Is it > > > reasonable to continue someone else's work when they've disappeared? > > > > > > Jon > > > > > > > > > -- > > Jonathan Ellis > > co-founder, http://www.datastax.com > > @spyced > > - To unsubscribe, e-mail: dev-unsubscr...@cassandra.apache.org For additional commands, e-mail: dev-h...@cassandra.apache.org
Re: server side describe
> > Someone once said: In my opinion, sniping like this doesn't help us move the conversation forward. Please reach out to other contributors who's behavior you have concerns with separately. On Fri, Apr 3, 2020 at 12:23 PM Joshua McKenzie wrote: > This isn't a hill to die on or something to binding -1 for me > personally. In a vacuum this merge is totally fine. The problem for me > comes in if a merge like this is one of 10, or 50, or 100 things that > are innocuous in isolation. IMO as long as we make sure this is the > only cut we do to ourselves we won't face death by a thousand cuts on > 4.0. > > In general I'm concerned about our rigor and discipline on restricting > scope to get 4.0 out the door, but that's in no way unique to us as a > project or 4.0 as a release; this has happened with every large > software release I've ever worked on and always requires significant > discomfort to lock down. It's unfortunate that the situation with this > ticket stumbled across that sore point for me (and likely many of us), > but I think we should keep that to the other thread about scope. > > On Fri, Apr 3, 2020 at 11:34 AM Jeff Jirsa wrote: > > > > Someone once said: > > > > "I heard the expression recently that “there are ten ways to do this, and > > eight of them will work.” I think that applies to most of the code we > > write. We don't need to spend a lot of time discussing which of the > eight > > is best; let’s trust the judgement of the original author and move > > forward. " > > > > Had we applied that principle to this JIRA the first time the patch was > > available in OCTOBER 2018, we wouldn't be having a conversation about > > whether or not it violates a freeze. > > > > > > On Wed, Apr 1, 2020 at 11:37 AM Jonathan Ellis > wrote: > > > > > I think we should get serious about the so-called freeze. > > > > > > On Wed, Apr 1, 2020 at 1:27 PM Jon Haddad wrote: > > > > > > > Hey folks, > > > > > > > > I was looking through our open JIRAs and realized we hadn't merged in > > > > server side describe calls yet. The ticket died off a ways ago, and > I > > > > pinged Chris about it yesterday. He's got a lot of his plate and > won't > > > be > > > > able to work on it anytime soon. I still think we should include > this in > > > > 4.0. > > > > > > > > From a technical standpoint, It doesn't say much on the ticket after > > > Robert > > > > tossed an alternative patch out there. I don't mind reviewing and > > > merging > > > > either of them, it sounded like both are pretty close to done and I > think > > > > from the perspective of updating drivers for 4.0 this will save > quite a > > > bit > > > > of time since driver maintainers won't have to add new CQL > generation for > > > > the various new options that have recently appeared. > > > > > > > > Questions: > > > > > > > > * Does anyone have an objection to getting this into 4.0? The patches > > > > aren't too huge, I think they're low risk, and also fairly high > reward. > > > > * I don't have an opinion (yet) on Robert's patch vs Chris's, with > regard > > > > to which is preferable. > > > > * Since soon after Robert put up his PR he hasn't been around, at > least > > > as > > > > far as I've seen. How have we dealt with abandoned patches before? > If > > > > we're going to add this in the patch will need some cleanup. Is it > > > > reasonable to continue someone else's work when they've disappeared? > > > > > > > > Jon > > > > > > > > > > > > > -- > > > Jonathan Ellis > > > co-founder, http://www.datastax.com > > > @spyced > > > >
Re: server side describe
> scope creep. I think it is unfair to label this scope creep; it would have to be newly considered for 4.0 for it to fall under that umbrella. I don't personally mind if it lands, but this was discussed at length on multiple occasions over the past year, and only stalled because of a combination of lack of etiquette, and a lack of leadership from e.g. PMC in resolving various political questions over the course of events. I also struggle to see how this would invalidate testing in any significant way? It doesn't modify any existing behaviour. From: Joshua McKenzie Sent: 01 April 2020 19:24 To: dev@cassandra.apache.org Subject: Re: server side describe This looks like a feature that'd potentially invalidate some testing that's been done and we've been feature frozen for over a year and a half. Also: scope creep. My PoV is we hold off. If we get into a cadence of more frequent releases we'll have it soon enough. On Wed, Apr 1, 2020 at 3:03 PM wrote: > Hi, > Normally I ping the person on the ticket or in Slack to ask him/her for > status update and whether I can help. Then probably he/she gives me a > direction. > If I can’t find the person anymore, I just use my best judgement and > coordinate with people who might know better than me. > For now this strategy worked for me personally. > Hope this helps > Ekaterina > > Sent from my iPhone > > > On 1 Apr 2020, at 14:27, Jon Haddad wrote: > > > > Hey folks, > > > > I was looking through our open JIRAs and realized we hadn't merged in > > server side describe calls yet. The ticket died off a ways ago, and I > > pinged Chris about it yesterday. He's got a lot of his plate and won't > be > > able to work on it anytime soon. I still think we should include this in > > 4.0. > > > > From a technical standpoint, It doesn't say much on the ticket after > Robert > > tossed an alternative patch out there. I don't mind reviewing and > merging > > either of them, it sounded like both are pretty close to done and I think > > from the perspective of updating drivers for 4.0 this will save quite a > bit > > of time since driver maintainers won't have to add new CQL generation for > > the various new options that have recently appeared. > > > > Questions: > > > > * Does anyone have an objection to getting this into 4.0? The patches > > aren't too huge, I think they're low risk, and also fairly high reward. > > * I don't have an opinion (yet) on Robert's patch vs Chris's, with regard > > to which is preferable. > > * Since soon after Robert put up his PR he hasn't been around, at least > as > > far as I've seen. How have we dealt with abandoned patches before? If > > we're going to add this in the patch will need some cleanup. Is it > > reasonable to continue someone else's work when they've disappeared? > > > > Jon > > - > To unsubscribe, e-mail: dev-unsubscr...@cassandra.apache.org > For additional commands, e-mail: dev-h...@cassandra.apache.org > >
