Closed m-kuhn closed 7 years ago
A bit scary ;) What does it mean?
rm -rf src/server
rm remove
- watch out: here come options!!
r recursively
f force
src/ in the sourcecode
server the server
An alternative approach of someone taking responsibility for the code to migrate/refactor/maintain could be discussed ;)
Thank you for the explanation, my question was not really about the syntax of rm
but about the motivation behind the removal of server from QGIS source.
Do you mean that QGIS Server is going to be removed from the QGIS project?
I might have missed the thread but I've not seen much discussion on this topic in the mailing lists.
Mathias may you rephrase more explicitly the issue title and motivations. A good solution is to close this semi-cryptic issue, and open it again with a title that can warn correctly qgis server users and devs.
On a more serious note, no, that's not directly meant. It's not going to be removed.
What's meant is that QGIS server is often neglected and in it's current state not sustainable. It's in need for a major refactoring and for a fully dedicated maintainer!
We should open a QEP and make an audit of what is good and what is wrong in current status. Many users rely heavily on QGIS Server, so we do not want to scare them. The goal is the refactor the server to make it sustainable
@m-kuhn LOL, can somenone add following tags :+1:
I fully agree that the server deserves more love. BTW I wouldn't say that the server is unmaintained, there is a small group of devs that is constantly taking care of it by backporting bug fixes, improving documentation etc. I'm pretty sure that we can create a team to ensure QGIS Server a bright future in the 3.0 branch.
@elpaso I know and it is much appreciated!
Does the server need to be in the same repo with the core or could it make sense to give it its own release cycle ?
I would prefer to make it shiny and leave it where it is as a first-class-citizen.
we could redevelop it in pure Python and use QGIS api, and should get rid of old fcgi dependancy. We could develop a Python API to implement the OGC interface
I can totally understand the frustration with the extra maintenance burden that server brings. It can be very frustrating when an otherwise well designed change in qgis core breaks something in server because server is misusing parts of the qgis api. To keep the lighthearted tone of this thread... whenever I see how server is handling project loading and the oddness it does with layer handling I can't help but think of this tweet:
This code looks pretty old, I wonder if I can delete th.... what the... pic.twitter.com/mUZzLbQzrG
— Merrick Christensen (@iammerrick) September 18, 2016
QGIS server has some fantastic features, and I'm really happy to see the increased attention being given to making it stable and covered by regression tests.
...but I get very nervous when I see things like the number of potential security issues flagged by coverity against the server components. Online security (and server uptime!) is critical stuff and unfortunately instead of being the most maintained and thoroughly tested component of qgis it's unfortunate that server is currently lagging far behind with the attention that core/gui/app/plugins get.
Speaking totally from my own viewpoint... I think there's a two reasons why this is:
How can we fix this? I'm not sure... I think my first point above will always stand, but point 2 could be aided by further pushing the message that QGIS' survival is 100% dependent on people giving back to the community and not leaching off it.
If there are companies behind, it's a pity we didn't see a grant proposal for this then.
Matthias' solution based on rm
could be a good approach. Wait for a new implementation and ship 3.0 with or without server.
Hi alls,
Thanks Mathias to initiate this discussion !
A) Short answer:
Some key features on QGIS server should be kept, but the code and it’s design not. Means: rm -rf server/* + git pulls
B) Spirit:
The feature to keep is the WMS server based on .qgs symbology.
Ability for an end-user to deal with a rich symbology trough a GUI, and to get the really same output both with his desktop app and on the Web, worth. A lot.
And for same reason, as an extra WMS feature, PDF output (GetPrint), should also be kept.
On the other side, all the others features are not that meaningful, as they could already been provided (and better) by some others apps (WFS, TMS, SOS, Auth…)
So IMO a good target for 3.0 could/should be: WMS + SLD + FE + GetPrint PDF support
C) Few Architecture thoughts:
OGC support must be exhaustive And so pass all related OGC CITE tests If not the solution just miss interoperability point
If possible this implementation should be done appart (no Qt dependancy for instance) to be reusable in others projects.
D) Oslandia could help on this QGIS Server 3.0 code effort.
Reactions welcome :)
@mdouchin this is already possible, QGIS Server is built as a library and FCGI is just a tiny wrapper around QgsServer
instance.
We should not drop all features not related to WMS. As it is very good to style layers in QGIS and have great WMS images with rich QGIS rendering, it is also very handy to configure the WFS within QGIS : fields to publish, aliases, use virtual fields with expressions, add filters via expressions, etc. QGIS helps to configure layers options, not only symbology.
Please do not drop features too fast. Many users rely on them now.
@ocourtin I agree that Auth is out of scope for the QGIS Server tier, but I'd like to keep all the main OGC services available, what I was thinking of is a C++ plugin-based architecture in which every service is a independent plugin, this would make the transition much smoother:
+1 for keeping at least WFS. As for dropping features, we must be careful: if QGIS3 will not have an usable server, people will be forced to stick to QGIS2, or to keep both version, thus resulting in an effective forking of the project, or at least of the community, a Bad Thing that caused huge problems in several important OS projects.
@m-kuhn @nyalldawson what can we do to fix the QGIS Server tests ? https://github.com/qgis/QGIS/pull/3510 What can we have to do to have Qt5 and Python3 well support in QGIS Server ?
Why not spit QGIS Server off as it's own project and repo? Does it make sense to keep in with the desktop source?
On Fri, Sep 23, 2016 at 9:53 PM, rldhont notifications@github.com wrote:
@m-kuhn https://github.com/m-kuhn @nyalldawson https://github.com/nyalldawson what can we do to fix the QGIS Server tests ? qgis/QGIS#3510 https://github.com/qgis/QGIS/pull/3510 What can we have to do to have Qt5 and Python3 well support in QGIS Server ?
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/qgis/qgis3.0_api/issues/63#issuecomment-249172493, or mute the thread https://github.com/notifications/unsubscribe-auth/AAXS3GAqDxjL1YRbpd0MEP_QjZodh_G-ks5qs73VgaJpZM4KDndY .
@NathanW2 I think it does, as long as they share the same project files they need to be in sync. IMHO the unique selling point of QGIS Server is that you create a project on the desktop and publish it on the server and see the same rendering on both.
Yeah not saying that just from a code point of view putting it outside and managing as a standalone project. I don't use it so I have no input on it's direction into the future.
On Fri, Sep 23, 2016 at 10:17 PM, Alessandro Pasotti < notifications@github.com> wrote:
@NathanW2 https://github.com/NathanW2 I think it does, as long as they share the same project files the need to be in sync. IMHO the unique selling point of QGIS Server is that you create a project on the desktop and publish it on the server and see the same rendering on both.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/qgis/qgis3.0_api/issues/63#issuecomment-249176774, or mute the thread https://github.com/notifications/unsubscribe-auth/AAXS3EqvVaLRd2I_FvPR5qZdK3glkcXpks5qs8NTgaJpZM4KDndY .
@elpaso @mdouchin @pcav
WFS implementation (i mean an interoperable one) is something big to achieve (and then to maintain).
It would be that cheaper for QGIS project to just provide kind of binding allowing from a .qgs other already existing WFS server to deal with.
On this way you don't loose feature for the end user. In fact you just provide more and better...
On Fri, Sep 23, 2016 at 05:17:23AM -0700, Alessandro Pasotti wrote:
@NathanW2 I think it does, as long as they share the same project files the need to be in sync. IMHO the unique selling point of QGIS Server is that you create a project on the desktop and publish it on the server and see the same rendering on both.
This reminds me of another discussion about having a formal specification for the project files.
Can I suggest we close this issue - I think if we table ideas like this it would be better to do it in a more constructive way. A lot of people use QGIS Server and I do not think we would serve them well by simply removing the code for QGIS server from the code base. Thanks!
Happy to close it. I has triggered a very constructive discussion. Long live QGIS server :+1:
QGIS Server is not going to be removed but it's in strong need for a dedicated maintainer!