Re: [QGIS-Developer] [Qgis-user] Thoughts on QGIS Development and LTR Releases
On Mon, 2 Mar 2020 at 05:03, Micha Silver wrote: > > This discussion has raised important issues and clarified concerns from the > user's perspective. But there is another hurdle that lies between developers > and users that hasn't been mentioned yet (not in this thread, anyway): > packaging. > > > Different packagers (on different repositories) release versions on different > schedule that may or may not be synchronized with the developers. What makes > the situation even more troublesome is release of dependencies at yet another > schedule. So all too often we find ourselves stuck after a routine system > update, with unresolved dependencies and partially or totally unusable > software. This has happened with QGIS and the grass plugin, and recently with > updates to proj4 and gdal. > > > There's no doubt that moving to a 2 year life cycle for LTS will help > mitigate this recurring problem. Actually, a 2 year cycle will make this far, far more problematic. An example is the proj library. QGIS 3.4 was developed using the older proj v4 library, and while it's possible to build it using proj v 6 or v7, it's not far from recommended. When proj v8 is released then it will become completely impossible to build QGIS 3.4. If we tried to squeeze another 12 months out of QGIS 3.4 then we'll bump into all sorts of linux packaging issues as the underlying linux distros upgrade to proj v6 or later. And that's completely out of our control... Nyall ___ QGIS-Developer mailing list QGIS-Developer@lists.osgeo.org List info: https://lists.osgeo.org/mailman/listinfo/qgis-developer Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-developer
Re: [QGIS-Developer] [Qgis-user] Thoughts on QGIS Development and LTR Releases
This discussion has raised important issues and clarified concerns from the user's perspective. But there is another hurdle that lies between developers and users that hasn't been mentioned yet (not in this thread, anyway): packaging. Different packagers (on different repositories) release versions on different schedule that may or may not be synchronized with the developers. What makes the situation even more troublesome is release of dependencies at yet another schedule. So all too often we find ourselves stuck after a routine system update, with unresolved dependencies and partially or totally unusable software. This has happened with QGIS and the grass plugin, and recently with updates to proj4 and gdal. There's no doubt that moving to a 2 year life cycle for LTS will help mitigate this recurring problem. On 01/03/2020 20:02, Brent Wood wrote: I'd also like to toss in my preference for a 2 year LTS. My reasons are similar to others who have al;ready posted. I've been using QGIS since v0.2, and running QGIS training workshops annually for several years now. While the new capabilities are great, I find users are frustrated in that they have difficulty keeping up - no sooner do they become comfortable with a version & they need to upgrade. People coming on my workshops are running various versions, and so some things work & others don't. I run 3.4, 3.10 (& still 2.18) as well as the nightly builds at times... so appreciate the benefits & issues of the current situation. IMHO, the extra stability of a longer term LTS will, overall, be beneficial. Thanks Brent Wood Brent Wood Programme leader: Environmental Information Delivery NIWA DDI: +64 (4) 3860529 From: Qgis-user on behalf of Ben Hur Pintor Sent: Sunday, March 1, 2020 17:41 To: Alexandre Neto Cc: QGIS User; qgis-developer Subject: Re: [Qgis-user] [QGIS-Developer] Thoughts on QGIS Development and LTR Releases Hi everyone, I'm not a QGIS developer but I've been a user since 1.X days. Aside from a few quirks (e.g. dependency issues on Linux), I've always liked the QGIS release cycle so I'm coming from that perspective. To add to Alexandre's explanation, this is usually how I explain QGIS releases: There are 3 branches of development - LR (Latest Release), LTR (Long Term Release), and nightly builds. Let's take a look at the LR and LTR branches. I usually take "maintained" to mean that the branch is actively developed -- bugs will be fixed, new features are added, etc. The LTR (now 3.10.3) is maintained until the next LTR (in 12 months when 3.16.4 becomes the LTR). During those 12 months, a PR (Point Release) will be released for the LTR branch each month (named 3.10.4, 3.10.5, ..., 3.10.14). This PR includes the bug fixes for 3.10.X. Being an LTR means that there (usually) won't be any changes that will break that version in 1 year. This is usually why people think of LTR as "stable". The LR (now 3.12.0) contains the most recent features of QGIS. A new LR is released every 4 months. The next LR (3.14) is slated to be released in June 2020, the next after that is 3.16 in October 2020. The 2nd LR released after the release of the LTR is the version that will become the next LTR. In this case, the 3.16 version that will be released in October 2020 is slated to become the LTR after 4 months (e.g. 3.16.4 will become the LTR by February 2021). For each month, a PR is also released for the LR branch until a new LR is released. For example, there will be a 3.12.1, 3.12.2, and 3.12.3 releases for March, April, and May 2020 until the 3.14 release in June. This repeats for 3.14.1,..,3.14.3 until the 3.16 release in October. Throughout the year, there can be big changes between LR versions (e.g. between 3.12, 3.14, and 3.16) as compared to the LTR which stays at 3.10.X. It's also worth noting that useful features in the LR branch can be backported (added to) the LTR branch. The "LR/PR" released every 4 months denotes that the release is a new LR. The "LTR/PR" released in October denotes that this release will be the next LTR. It is not put in the LTR repo until 4 months later (February) where it officially becomes the LTR. We can also think of it this way: For the LTR branch, you can expect updates every month of its life (or with every point release) with 3.10.3, 3.10.4, 3.10.5, ..., 3.10.14 but these will usually be minor and not so drastic. Meanwhile, there may be drastic and major changes between LRs like 3.12, 3.14, 3.16, etc. but only minor changes between PRs of the same LR -- eg 3.12.1 and 3.12.2, 3.14.1 and 3.14.2, etc. I also agree that the road map found at https://www.qgis.org/en/site/getinvolved/development/roadmap.html is geared a lot towards dev people but, in its defense,
Re: [QGIS-Developer] [Qgis-user] Thoughts on QGIS Development and LTR Releases
Hi everyone, I'm not a QGIS developer but I've been a user since 1.X days. Aside from a few quirks (e.g. dependency issues on Linux), I've always liked the QGIS release cycle so I'm coming from that perspective. To add to Alexandre's explanation, this is usually how I explain QGIS releases: There are 3 branches of development - LR (Latest Release), LTR (Long Term Release), and nightly builds. Let's take a look at the LR and LTR branches. I usually take "maintained" to mean that the branch is actively developed -- bugs will be fixed, new features are added, etc. The LTR (now 3.10.3) is maintained until the next LTR (in 12 months when 3.16.4 becomes the LTR). During those 12 months, a PR (Point Release) will be released for the LTR branch each month (named 3.10.4, 3.10.5, ..., 3.10.14). This PR includes the bug fixes for 3.10.X. Being an LTR means that there (usually) won't be any changes that will break that version in 1 year. This is usually why people think of LTR as "stable". The LR (now 3.12.0) contains the most recent features of QGIS. A new LR is released every 4 months. The next LR (3.14) is slated to be released in June 2020, the next after that is 3.16 in October 2020. The 2nd LR released after the release of the LTR is the version that will become the next LTR. In this case, the 3.16 version that will be released in October 2020 is slated to become the LTR after 4 months (e.g. 3.16.4 will become the LTR by February 2021). For each month, a PR is also released for the LR branch until a new LR is released. For example, there will be a 3.12.1, 3.12.2, and 3.12.3 releases for March, April, and May 2020 until the 3.14 release in June. This repeats for 3.14.1,..,3.14.3 until the 3.16 release in October. Throughout the year, there can be big changes between LR versions (e.g. between 3.12, 3.14, and 3.16) as compared to the LTR which stays at 3.10.X. It's also worth noting that useful features in the LR branch can be backported (added to) the LTR branch. The "LR/PR" released every 4 months denotes that the release is a new LR. The "LTR/PR" released in October denotes that this release will be the next LTR. It is not put in the LTR repo until 4 months later (February) where it officially becomes the LTR. We can also think of it this way: For the LTR branch, you can expect updates every month of its life (or with every point release) with 3.10.3, 3.10.4, 3.10.5, ..., 3.10.14 but these will usually be minor and not so drastic. Meanwhile, there may be drastic and major changes between LRs like 3.12, 3.14, 3.16, etc. but only minor changes between PRs of the same LR -- eg 3.12.1 and 3.12.2, 3.14.1 and 3.14.2, etc. I also agree that the road map found at https://www.qgis.org/en/site/getinvolved/development/roadmap.html is geared a lot towards dev people but, in its defense, it is located under the "Get Involved/Development" part of the documentation. If I have misrepresented or misunderstood anything, please don't hesitate to correct me. All the best, Ben Hur On Sun, Mar 1, 2020 at 12:13 PM Alexandre Neto wrote: > Hi Groene, > > I agree that the Road map is not easy to understand. Just to clarify > things (I hope): > > The current LTR is 3.10 (currently at 3.10.3). It only became LTR in > February although its first release was done in october (3.10.0). The idea > is to let it mature (and have a broader usage and tests) for at least 4 > months before it becomes LTR. > > This means that for the next 12 months, more or less every month a new > patch release will come out. That will be 3.10.4, 3.10.5 and so on. Only > bug fixes are allowed on these releases, no new features. > > If there is a version with some extra number (e.g. 3.4.13-3) it means that > something happened during packaging of the patch release, requiring a new > package/installer to be created. You should always try to use the latest > version of the LTR version, which should be the most stable one. > > Hope it helps. > > Alexandre Neto > > On Sat, Feb 29, 2020 at 7:11 PM Groene Bij wrote: > >> Hi all, >> >> >> >> Thank you for bringing up this topic. >> >> For me the release schedule and the different abbreviations are not very >> helpful when choosing a qgis version to install. I am not a software >> developer and thus have little understanding what the different releases >> are all about. >> >> >> >> Clarity, however, is always appreciated, and that is the main thing >> missing in this topic: >> >> >> >> Chris is talking about 3.10.2 having been labeled LTR >> >> Qgis.org right now (29th of February) is mentioning 3.10.3 as LTR >> >> The release schedule on qgis.org is mentioning 3.10.3 as LR with 3.4.13 >> being the most recent LTR >> >> So things are unclear. Qgis.org itself seems to be unclear which version >> actually is the LTR. >> >> >> >> The information on the Road Map page (schedule release) is also unclear: >> >> “The schedule is aligned to produce roughly the same dates for each year >> given our four monthly releases
Re: [QGIS-Developer] [Qgis-user] Thoughts on QGIS Development and LTR Releases
Hi Groene, I agree that the Road map is not easy to understand. Just to clarify things (I hope): The current LTR is 3.10 (currently at 3.10.3). It only became LTR in February although its first release was done in october (3.10.0). The idea is to let it mature (and have a broader usage and tests) for at least 4 months before it becomes LTR. This means that for the next 12 months, more or less every month a new patch release will come out. That will be 3.10.4, 3.10.5 and so on. Only bug fixes are allowed on these releases, no new features. If there is a version with some extra number (e.g. 3.4.13-3) it means that something happened during packaging of the patch release, requiring a new package/installer to be created. You should always try to use the latest version of the LTR version, which should be the most stable one. Hope it helps. Alexandre Neto On Sat, Feb 29, 2020 at 7:11 PM Groene Bij wrote: > Hi all, > > > > Thank you for bringing up this topic. > > For me the release schedule and the different abbreviations are not very > helpful when choosing a qgis version to install. I am not a software > developer and thus have little understanding what the different releases > are all about. > > > > Clarity, however, is always appreciated, and that is the main thing > missing in this topic: > > > > Chris is talking about 3.10.2 having been labeled LTR > > Qgis.org right now (29th of February) is mentioning 3.10.3 as LTR > > The release schedule on qgis.org is mentioning 3.10.3 as LR with 3.4.13 > being the most recent LTR > > So things are unclear. Qgis.org itself seems to be unclear which version > actually is the LTR. > > > > The information on the Road Map page (schedule release) is also unclear: > > “The schedule is aligned to produce roughly the same dates for each year > given our four monthly releases with LTRs in late february.” > > Looking at the schedule, LTR’s are being released in October, not February. > > > > > > In this mail Régis says: > > “LTR does not mean stable. LTR means it will gain bugfixes longer than > releases.” > > What does this actually mean? Does this mean 3.4.13 will have releases > like 3.4.13-1 and -2 etc.? I do actually find them in the download section > ‘older releases’ or ‘previous releases’ (also no idea why there are two > different download sections). > > I do not find any information on the release schedule page about LTR’s > having different bugfix releases. The Road Map page only says: ´ Every > third release (starting with 2.8) is a long-term-release (LTR) *that is > maintained* until the next long-term-release occurs. To me, ‘maintained’ > can mean a lot of things. > > > > Considering LTR and PR: > > What exactly is the difference between 3.4.13 and 3.4.14? Does 3.4.14 > contain the same bugfixes as 3.4.13-3? The most recent release date of > 3.4.14 is 2019-dec-07 and for 3.4.13-3 is 2019-dec-05. > > Or does 3.4.14 only contain new features, compared to the original 3.4.13, > released in October? > > Additional information on how to ‘read’ the release schedule would be much > appreciated. I find the current information difficult to understand and it > seems to me it is written too much from a developers or IT-minded > perspective in stead od a general user perspective. > > > > Best regards, > > Jeroen Hovens > > > > > > *Van:* Qgis-user *Namens *Régis > Haubourg > *Verzonden:* donderdag 20 februari 2020 18:25 > *Aan:* C Hamilton > *CC:* qgis-user ; qgis-developer < > qgis-developer@lists.osgeo.org> > *Onderwerp:* Re: [Qgis-user] Thoughts on QGIS Development and LTR Releases > > > > Hi Chris, > > I share most of your concerns, as much as I advocate the spread of QGIS in > enterprise and organisations. > > It is true we need always more reliability, documentation. I'd like also > to point that 2.x is not so far away, and that the reliability have since > improved by order of magnitude. > > Let's also keep in mind that the level of expectations of users grows very > fast too, so this is a race that will never end ;-) > > > > However, I think there is a cultural problem, and probably a pedagogy > effort we should make. > > > > LTR does not mean stable. LTR means it will gain bugfixes longer than > releases. So it is highly expectable that installing a LTR in its early > versions will let you hit more issues. I remember the very same situation > for ArcGIS 8 or 9 early stages. And this is the very same for linux > distributions or any software. I don't remember any early x.0 release in > QGIS that was not followed one week later by an urgent point release. But > new users don't know this. They see a big green button "download that sexy > new version". > > > > That said, how to improve the situation? After years of discussions in the > various events, hackfest, conferences, discussions with public or private > customers, developpers, here are the possible leads we have: > > > > - Keep on explaining the rationale and codes of free software to users and >
Re: [QGIS-Developer] [Qgis-user] Thoughts on QGIS Development and LTR Releases
Hi Chris, I share most of your concerns, as much as I advocate the spread of QGIS in enterprise and organisations. It is true we need always more reliability, documentation. I'd like also to point that 2.x is not so far away, and that the reliability have since improved by order of magnitude. Let's also keep in mind that the level of expectations of users grows very fast too, so this is a race that will never end ;-) However, I think there is a cultural problem, and probably a pedagogy effort we should make. LTR does not mean stable. LTR means it will gain bugfixes longer than releases. So it is highly expectable that installing a LTR in its early versions will let you hit more issues. I remember the very same situation for ArcGIS 8 or 9 early stages. And this is the very same for linux distributions or any software. I don't remember any early x.0 release in QGIS that was not followed one week later by an urgent point release. But new users don't know this. They see a big green button "download that sexy new version". That said, how to improve the situation? After years of discussions in the various events, hackfest, conferences, discussions with public or private customers, developpers, here are the possible leads we have: - Keep on explaining the rationale and codes of free software to users and potential funders. - Try to keep our "power users / early testers" population, so that we target the right issues during bugfix sprints. - Offer longer LTR lifespan, so that funders have a larger window to actually find and have bug fixed. - Keep on explaining that QGIS bugfix release should be easily deployable in big organisations. OSGEO4W silent installs allows this. Maybe going toward auto upgrade / patch system could help (it's a big effort though) - Keep on gaining more budget for QGIS.org, so that we can setup a real semi automated Q/A acceptance test suite. This requires human tests. Boundless did, it is possible. It is a matter of ressources. Should it be centralized or community powered ? I have no idea, but this requires someone to be hired all year long to do this. IMO, enterprises requiring such reliability should really consider sponsoring this framework and dedicate some human ressources. - Same goes for documentation - Same goes for code review, we need to have more reviewers. the learning curve is steep though, and we need to find money for this - Improve the website with a simple page, with graphics and videos on what is the lifecycle of QGIS, and what version to use for what expectations. A note about QGIS.org budget. To me, it is only a leverage, a catalyser, but it can't fund itself a full QA infrastructure with the current economic model of the association. I think, this is our responsability to spread this word everywhere so that the user / contributor rate changes a bit. After all, even Microsoft with its thousands of testers, and its early testing network was able to push updates causing the famous Blue Screen Of the Death. So shit can happen. Packaging nightmare with major changes in underlying libraries remains a really really complex process. How fast we are to fix and change our ways to do is the real question. I think the QGIS and OSGeo Community does a tremendeous work. Best regards, Régis Le jeu. 20 févr. 2020 à 16:21, C Hamilton a écrit : > I first want to say how much I appreciate all of the QGIS developers and > all of your hard work, but I would also like to suggest that you exercise > caution when you label a release LTR. I work in a large organization where > most geospatial analysts can have access to ArcGIS if they want it. The > advantage to ArcGIS is that everyone has been trained to use it, ESRI has > been around for a long time and there is a lot of documentation, training > and support for it. So why would users want to use QGIS? > > There are always a curious few who see QGIS and realize they can download > it for free at home. They tinker with it and come to like it and then they > try it in the workplace. For the users who have ArcGIS at their disposal > there must be a good reason to use QGIS instead. These tend to be the > reasons they use QGIS: 1) It does not crash as much as ArcGIS. 2) It is > faster than ArcGIS. 3) It can effectively processing larger data sets than > ArcGIS. 4) There may be some workflow in QGIS that is simpler than in > ArcGIS. > > I think that the QGIS community can be proud about the fact that most of > my users who start using QGIS love it and don't want to go back to ArcGIS > if at all possible. > > If a user finds that their reason for using QGIS goes away, they will be > disappointed, but will to go back to ArcGIS. I am an advocate for QGIS in > our work place. I think it should be used more, but it is really, really > hard to convince most people. Most of my users are not programmers so if > something is broken they don' t know how to fix it. We have QGIS support > contracts which help. Users consider the QGIS
