[webkit-dev] What is an "active port"? [WAS: Do you maintain OS(WINCE)?]
Hi Patrick. > On Tue, Sep 13, 2011 at 2:38 PM, Patrick Gansterer wrote: > How do we measure an "active port"??? I maintain a buildbot for WinCe and > usually fix problems with the port within hours. Unfortunately I don't get > paid to work on WebKit the whole day and so I can't make such big steps > forward like other ports do. In my effort to establish the "threads exist" baseline abstraction, I've gotten a few responses similar to this one: "I maintain port X, but I'm the only one, and I have limited time…". Here are my current thoughts, based on that experience: * A long list of #ifdefs in core WebKit code makes reading and understanding the code difficult, especially if the #ifdefs select among a matrix of fundamentally different ways of doings things. * A long tail of ports makes fundamental improvements to the engine difficult and time consuming. Fundamental improvements are likely to break a port, and port maintainers may not be available in a timely fashion to adopt a fundamental improvement. (For example, it has been about a week since I started the "threads exist" project.) * We have a significant number of ports (maybe 5) that are either (a) maintained by only one person working part-time or (b) not maintained at all in WebKit trunk, but periodically upstreamed to WebKit trunk by downstream clients to make their future merges easier. * Single-part-time-maintainer ports seem to keep up at a reasonable pace with simple build fixes like adding new files to projects, but not with major architectural changes. * Single-part-time-maintainer ports get very little, if any, testing outside of automated regression tests, so it's hard to know if the code actually works, who uses it, or what its requirements are. When I ask if a port is "active", I guess what I mean is, "Can I go ahead and make this core WebKit improvement, and assume that port maintainers will keep up, or do I need to stop what I'm doing and wait for them, or worry that they will roll out some or all of my patch instead of doing the harder work of upgrading their port?" I also mean, "Is this port actively used, and is the opportunity cost of upgrading it justified?" I think the right solution here is for port maintainers, in cases of nontrivial work, to take on the job of upgrading their ports to match core engine changes, instead of core engineers taking on that job. And, in cases where a port upgrade isn't available in a timely fashion for some reason, WebKit should move forward and break the port (core builder or not). This proposal might seem unkind, but I think it's the best thing for moving WebKit forward in the long run. > On Tue, Sep 13, 2011 at 2:38 PM, Patrick Gansterer wrote: > So PLEASE: When do we call a port "active"? It's not cool to get the question > about removal every few months! I hope that the plan I've outlined above will make "active" ports much more well-known to core WebKit contributors, since port maintainers will be working with core contributors to upgrade their ports. Regards, Geoff___ webkit-dev mailing list webkit-dev@lists.webkit.org http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
Re: [webkit-dev] What is an "active port"? [WAS: Do you maintain OS(WINCE)?]
One of the things I admire about the WebKit project is that historically the project has been very inclusive. One common thread that's woven through a number of recent discussions is that folks feel we've taken on too much complexity and that it's harder to make fundamental improvements to the engine. Maybe it's time to adjust that balance slightly, but I hope we adjust our behavior deliberately. Adam On Wed, Sep 14, 2011 at 12:08 PM, Geoffrey Garen wrote: > Hi Patrick. > > On Tue, Sep 13, 2011 at 2:38 PM, Patrick Gansterer > wrote: > > How do we measure an "active port"??? I maintain a buildbot for WinCe and > usually fix problems with the port within hours. Unfortunately I don't get > paid to work on WebKit the whole day and so I can't make such big steps > forward like other ports do. > > In my effort to establish the "threads exist" baseline abstraction, I've > gotten a few responses similar to this one: "I maintain port X, but I'm the > only one, and I have limited time…". > Here are my current thoughts, based on that experience: > * A long list of #ifdefs in core WebKit code makes reading and understanding > the code difficult, especially if the #ifdefs select among a matrix of > fundamentally different ways of doings things. > * A long tail of ports makes fundamental improvements to the engine > difficult and time consuming. Fundamental improvements are likely to break a > port, and port maintainers may not be available in a timely fashion to adopt > a fundamental improvement. (For example, it has been about a week since I > started the "threads exist" project.) > * We have a significant number of ports (maybe 5) that are either (a) > maintained by only one person working part-time or (b) not maintained at all > in WebKit trunk, but periodically upstreamed to WebKit trunk by downstream > clients to make their future merges easier. > * Single-part-time-maintainer ports seem to keep up at a reasonable pace > with simple build fixes like adding new files to projects, but not with > major architectural changes. > * Single-part-time-maintainer ports get very little, if any, testing outside > of automated regression tests, so it's hard to know if the code actually > works, who uses it, or what its requirements are. > When I ask if a port is "active", I guess what I mean is, "Can I go ahead > and make this core WebKit improvement, and assume that port maintainers will > keep up, or do I need to stop what I'm doing and wait for them, or worry > that they will roll out some or all of my patch instead of doing the harder > work of upgrading their port?" I also mean, "Is this port actively used, and > is the opportunity cost of upgrading it justified?" > I think the right solution here is for port maintainers, in cases of > nontrivial work, to take on the job of upgrading their ports to match core > engine changes, instead of core engineers taking on that job. And, in cases > where a port upgrade isn't available in a timely fashion for some reason, > WebKit should move forward and break the port (core builder or not). This > proposal might seem unkind, but I think it's the best thing for moving > WebKit forward in the long run. > > On Tue, Sep 13, 2011 at 2:38 PM, Patrick Gansterer > wrote: > > So PLEASE: When do we call a port "active"? It's not cool to get the > question about removal every few months! > > I hope that the plan I've outlined above will make "active" ports much more > well-known to core WebKit contributors, since port maintainers will be > working with core contributors to upgrade their ports. > Regards, > Geoff > ___ > webkit-dev mailing list > webkit-dev@lists.webkit.org > http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev > > ___ webkit-dev mailing list webkit-dev@lists.webkit.org http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
Re: [webkit-dev] What is an "active port"? [WAS: Do you maintain OS(WINCE)?]
Hi, I completely agree with all of your points. I also don't think that it's your task to keep every "part time" port working with every change. IMO most of the "is active" questions come with a "when do we remove the old code/port from trunk" question. That's not very cool to hear after the "hard" upstreaming work. But that's only my personal view. More interesting questing is: How do "part time" maintainers get "informed" about fundamental changes? I'd prefer cc'ing on a bug which might break a build. So it's possible for the maintainer to try to build with the patch locally and implement the missing parts. At least for me it's easier to fix compiler errors than answering questions about a possible build break on webkit-dev. ;-) - Patrick Am 14.09.2011 um 21:08 schrieb Geoffrey Garen: > Hi Patrick. > >> On Tue, Sep 13, 2011 at 2:38 PM, Patrick Gansterer wrote: >> How do we measure an "active port"??? I maintain a buildbot for WinCe and >> usually fix problems with the port within hours. Unfortunately I don't get >> paid to work on WebKit the whole day and so I can't make such big steps >> forward like other ports do. > > In my effort to establish the "threads exist" baseline abstraction, I've > gotten a few responses similar to this one: "I maintain port X, but I'm the > only one, and I have limited time…". > > Here are my current thoughts, based on that experience: > > * A long list of #ifdefs in core WebKit code makes reading and understanding > the code difficult, especially if the #ifdefs select among a matrix of > fundamentally different ways of doings things. > > * A long tail of ports makes fundamental improvements to the engine difficult > and time consuming. Fundamental improvements are likely to break a port, and > port maintainers may not be available in a timely fashion to adopt a > fundamental improvement. (For example, it has been about a week since I > started the "threads exist" project.) > > * We have a significant number of ports (maybe 5) that are either (a) > maintained by only one person working part-time or (b) not maintained at all > in WebKit trunk, but periodically upstreamed to WebKit trunk by downstream > clients to make their future merges easier. > > * Single-part-time-maintainer ports seem to keep up at a reasonable pace with > simple build fixes like adding new files to projects, but not with major > architectural changes. > > * Single-part-time-maintainer ports get very little, if any, testing outside > of automated regression tests, so it's hard to know if the code actually > works, who uses it, or what its requirements are. > > When I ask if a port is "active", I guess what I mean is, "Can I go ahead and > make this core WebKit improvement, and assume that port maintainers will keep > up, or do I need to stop what I'm doing and wait for them, or worry that they > will roll out some or all of my patch instead of doing the harder work of > upgrading their port?" I also mean, "Is this port actively used, and is the > opportunity cost of upgrading it justified?" > > I think the right solution here is for port maintainers, in cases of > nontrivial work, to take on the job of upgrading their ports to match core > engine changes, instead of core engineers taking on that job. And, in cases > where a port upgrade isn't available in a timely fashion for some reason, > WebKit should move forward and break the port (core builder or not). This > proposal might seem unkind, but I think it's the best thing for moving WebKit > forward in the long run. > >> On Tue, Sep 13, 2011 at 2:38 PM, Patrick Gansterer wrote: >> So PLEASE: When do we call a port "active"? It's not cool to get the >> question about removal every few months! > > I hope that the plan I've outlined above will make "active" ports much more > well-known to core WebKit contributors, since port maintainers will be > working with core contributors to upgrade their ports. > > Regards, > Geoff ___ webkit-dev mailing list webkit-dev@lists.webkit.org http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
Re: [webkit-dev] What is an "active port"? [WAS: Do you maintain OS(WINCE)?]
Maybe we need a webkit-port-maintainers@ list that one could easily cc rather than trying to add people by hand? -- Dirk On Wed, Sep 14, 2011 at 1:00 PM, Patrick Gansterer wrote: > Hi, > I completely agree with all of your points. I also don't think that it's > your task to keep every "part time" port working with every change. > IMO most of the "is active" questions come with a "when do we remove the old > code/port from trunk" question. That's not very cool to hear after the > "hard" upstreaming work. But that's only my personal view. > More interesting questing is: How do "part time" maintainers get "informed" > about fundamental changes? I'd prefer cc'ing on a bug which might break a > build. So it's possible for the maintainer to try to build with the patch > locally and implement the missing parts. At least for me it's easier to fix > compiler errors than answering questions about a possible build break on > webkit-dev. ;-) > - Patrick > Am 14.09.2011 um 21:08 schrieb Geoffrey Garen: > > Hi Patrick. > > On Tue, Sep 13, 2011 at 2:38 PM, Patrick Gansterer > wrote: > > How do we measure an "active port"??? I maintain a buildbot for WinCe and > usually fix problems with the port within hours. Unfortunately I don't get > paid to work on WebKit the whole day and so I can't make such big steps > forward like other ports do. > > In my effort to establish the "threads exist" baseline abstraction, I've > gotten a few responses similar to this one: "I maintain port X, but I'm the > only one, and I have limited time…". > Here are my current thoughts, based on that experience: > * A long list of #ifdefs in core WebKit code makes reading and understanding > the code difficult, especially if the #ifdefs select among a matrix of > fundamentally different ways of doings things. > * A long tail of ports makes fundamental improvements to the engine > difficult and time consuming. Fundamental improvements are likely to break a > port, and port maintainers may not be available in a timely fashion to adopt > a fundamental improvement. (For example, it has been about a week since I > started the "threads exist" project.) > * We have a significant number of ports (maybe 5) that are either (a) > maintained by only one person working part-time or (b) not maintained at all > in WebKit trunk, but periodically upstreamed to WebKit trunk by downstream > clients to make their future merges easier. > * Single-part-time-maintainer ports seem to keep up at a reasonable pace > with simple build fixes like adding new files to projects, but not with > major architectural changes. > * Single-part-time-maintainer ports get very little, if any, testing outside > of automated regression tests, so it's hard to know if the code actually > works, who uses it, or what its requirements are. > When I ask if a port is "active", I guess what I mean is, "Can I go ahead > and make this core WebKit improvement, and assume that port maintainers will > keep up, or do I need to stop what I'm doing and wait for them, or worry > that they will roll out some or all of my patch instead of doing the harder > work of upgrading their port?" I also mean, "Is this port actively used, and > is the opportunity cost of upgrading it justified?" > I think the right solution here is for port maintainers, in cases of > nontrivial work, to take on the job of upgrading their ports to match core > engine changes, instead of core engineers taking on that job. And, in cases > where a port upgrade isn't available in a timely fashion for some reason, > WebKit should move forward and break the port (core builder or not). This > proposal might seem unkind, but I think it's the best thing for moving > WebKit forward in the long run. > > On Tue, Sep 13, 2011 at 2:38 PM, Patrick Gansterer > wrote: > > So PLEASE: When do we call a port "active"? It's not cool to get the > question about removal every few months! > > I hope that the plan I've outlined above will make "active" ports much more > well-known to core WebKit contributors, since port maintainers will be > working with core contributors to upgrade their ports. > Regards, > Geoff > > ___ > webkit-dev mailing list > webkit-dev@lists.webkit.org > http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev > > ___ webkit-dev mailing list webkit-dev@lists.webkit.org http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
Re: [webkit-dev] What is an "active port"? [WAS: Do you maintain OS(WINCE)?]
To mitigate this issue, Leandro (acidx) and I are working on change log parser that can automatically detect active patch contributors and reviewers. (See https://bugs.webkit.org/show_bug.cgi?id=68061). Having said that, I think contributors should help maintaining ports that have bots on build.webkit.org or EWS bots. - Ryosuke On Wed, Sep 14, 2011 at 12:08 PM, Geoffrey Garen wrote: > In my effort to establish the "threads exist" baseline abstraction, I've > gotten a few responses similar to this one: "I maintain port X, but I'm the > only one, and I have limited time…". > > Here are my current thoughts, based on that experience: > > * A long list of #ifdefs in core WebKit code makes reading and > understanding the code difficult, especially if the #ifdefs select among a > matrix of fundamentally different ways of doings things. > > * A long tail of ports makes fundamental improvements to the engine > difficult and time consuming. Fundamental improvements are likely to break a > port, and port maintainers may not be available in a timely fashion to adopt > a fundamental improvement. (For example, it has been about a week since I > started the "threads exist" project.) > > * We have a significant number of ports (maybe 5) that are either (a) > maintained by only one person working part-time or (b) not maintained at all > in WebKit trunk, but periodically upstreamed to WebKit trunk by downstream > clients to make their future merges easier. > > * Single-part-time-maintainer ports seem to keep up at a reasonable pace > with simple build fixes like adding new files to projects, but not with > major architectural changes. > > * Single-part-time-maintainer ports get very little, if any, testing > outside of automated regression tests, so it's hard to know if the code > actually works, who uses it, or what its requirements are. > > When I ask if a port is "active", I guess what I mean is, "Can I go ahead > and make this core WebKit improvement, and assume that port maintainers will > keep up, or do I need to stop what I'm doing and wait for them, or worry > that they will roll out some or all of my patch instead of doing the harder > work of upgrading their port?" I also mean, "Is this port actively used, and > is the opportunity cost of upgrading it justified?" > > I think the right solution here is for port maintainers, in cases of > nontrivial work, to take on the job of upgrading their ports to match core > engine changes, instead of core engineers taking on that job. And, in cases > where a port upgrade isn't available in a timely fashion for some reason, > WebKit should move forward and break the port (core builder or not). This > proposal might seem unkind, but I think it's the best thing for moving > WebKit forward in the long run. > > On Tue, Sep 13, 2011 at 2:38 PM, Patrick Gansterer > wrote: > > So PLEASE: When do we call a port "active"? It's not cool to get the > question about removal every few months! > > > I hope that the plan I've outlined above will make "active" ports much more > well-known to core WebKit contributors, since port maintainers will be > working with core contributors to upgrade their ports. > > Regards, > Geoff > > ___ > webkit-dev mailing list > webkit-dev@lists.webkit.org > http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev > > ___ webkit-dev mailing list webkit-dev@lists.webkit.org http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
Re: [webkit-dev] What is an "active port"? [WAS: Do you maintain OS(WINCE)?]
For those of you interested in this stuff, I have a patch to add a webkit.org/team.html that auto-generates a list of contributors from committers.py on https://bugs.webkit.org/show_bug.cgi?id=68045. This version doesn't include area of expertise but we can add it easily once the bug 68061 is fixed (i.e. a script to detect active contributors and reviewers is added). - Ryosuke On Wed, Sep 14, 2011 at 1:08 PM, Ryosuke Niwa wrote: > To mitigate this issue, Leandro (acidx) and I are working on change log > parser that can automatically detect active patch contributors and > reviewers. (See https://bugs.webkit.org/show_bug.cgi?id=68061). > > Having said that, I think contributors should help maintaining ports that > have bots on build.webkit.org or EWS bots. > > - Ryosuke > > On Wed, Sep 14, 2011 at 12:08 PM, Geoffrey Garen wrote: > >> In my effort to establish the "threads exist" baseline abstraction, I've >> gotten a few responses similar to this one: "I maintain port X, but I'm the >> only one, and I have limited time…". >> >> Here are my current thoughts, based on that experience: >> >> * A long list of #ifdefs in core WebKit code makes reading and >> understanding the code difficult, especially if the #ifdefs select among a >> matrix of fundamentally different ways of doings things. >> >> * A long tail of ports makes fundamental improvements to the engine >> difficult and time consuming. Fundamental improvements are likely to break a >> port, and port maintainers may not be available in a timely fashion to adopt >> a fundamental improvement. (For example, it has been about a week since I >> started the "threads exist" project.) >> >> * We have a significant number of ports (maybe 5) that are either (a) >> maintained by only one person working part-time or (b) not maintained at all >> in WebKit trunk, but periodically upstreamed to WebKit trunk by downstream >> clients to make their future merges easier. >> >> * Single-part-time-maintainer ports seem to keep up at a reasonable pace >> with simple build fixes like adding new files to projects, but not with >> major architectural changes. >> >> * Single-part-time-maintainer ports get very little, if any, testing >> outside of automated regression tests, so it's hard to know if the code >> actually works, who uses it, or what its requirements are. >> >> When I ask if a port is "active", I guess what I mean is, "Can I go ahead >> and make this core WebKit improvement, and assume that port maintainers will >> keep up, or do I need to stop what I'm doing and wait for them, or worry >> that they will roll out some or all of my patch instead of doing the harder >> work of upgrading their port?" I also mean, "Is this port actively used, and >> is the opportunity cost of upgrading it justified?" >> >> I think the right solution here is for port maintainers, in cases of >> nontrivial work, to take on the job of upgrading their ports to match core >> engine changes, instead of core engineers taking on that job. And, in cases >> where a port upgrade isn't available in a timely fashion for some reason, >> WebKit should move forward and break the port (core builder or not). This >> proposal might seem unkind, but I think it's the best thing for moving >> WebKit forward in the long run. >> >> On Tue, Sep 13, 2011 at 2:38 PM, Patrick Gansterer >> wrote: >> >> So PLEASE: When do we call a port "active"? It's not cool to get the >> question about removal every few months! >> >> >> I hope that the plan I've outlined above will make "active" ports much >> more well-known to core WebKit contributors, since port maintainers will be >> working with core contributors to upgrade their ports. >> >> Regards, >> Geoff >> >> ___ >> webkit-dev mailing list >> webkit-dev@lists.webkit.org >> http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev >> >> > ___ webkit-dev mailing list webkit-dev@lists.webkit.org http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
Re: [webkit-dev] What is an "active port"? [WAS: Do you maintain OS(WINCE)?]
Responding to a few issues at once: On Sep 14, 2011, at 12:50 PM, Adam Barth wrote: > One of the things I admire about the WebKit project is that > historically the project has been very inclusive. One common thread > that's woven through a number of recent discussions is that folks feel > we've taken on too much complexity and that it's harder to make > fundamental improvements to the engine. Maybe it's time to adjust > that balance slightly, but I hope we adjust our behavior deliberately. Agreed. I think we can adjust toward putting more onus on port maintainers a step at a time, and see how things go. It may be a long-shot, but I really do believe that putting more onus on port maintainers has a chance to make WebKit *more* inclusive, not less, by bringing port maintainers into the fold of regular WebKit development. On Sep 14, 2011, at 1:00 PM, Patrick Gansterer wrote: > IMO most of the "is active" questions come with a "when do we remove the old > code/port from trunk" question. That's not very cool to hear after the "hard" > upstreaming work. But that's only my personal view. I'm sorry that you keep getting this question, especially from me. :) On Sep 14, 2011, at 1:00 PM, Patrick Gansterer wrote: > More interesting questing is: How do "part time" maintainers get "informed" > about fundamental changes? I'd prefer cc'ing on a bug which might break a > build. So it's possible for the maintainer to try to build with the patch > locally and implement the missing parts. At least for me it's easier to fix > compiler errors than answering questions about a possible build break on > webkit-dev. ;-) I think some combination of forewarning in Bugzilla bugs and email to webkit-dev is the right approach here. Hopefully, all active maintainers, by definition, keep up with webkit-dev email. Direct CC is a harder plan for me to work with, because I want to make sure that I get in touch will all interested port maintainers at once, even if I don't a priori know who they are. On Sep 14, 2011, at 1:02 PM, Dirk Pranke wrote: > Maybe we need a webkit-port-maintainers@ list that one could easily cc > rather than trying to add people by hand? Sounds helpful. Not sure exactly how it would work, though. (How would you add yourself to the list?) On Sep 14, 2011, at 1:08 PM, Ryosuke Niwa wrote: > Having said that, I think contributors should help maintaining ports that > have bots on build.webkit.org or EWS bots. To the extent that a bot can make certain kinds of fixes obvious, I think that core contributors should indeed make obvious fixes. However, I'm not convinced that having a bot should convey any special privileges beyond that. Geoff ___ webkit-dev mailing list webkit-dev@lists.webkit.org http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
Re: [webkit-dev] What is an "active port"? [WAS: Do you maintain OS(WINCE)?]
On Thu, Sep 15, 2011 at 12:17 PM, Geoffrey Garen wrote: > On Sep 14, 2011, at 1:02 PM, Dirk Pranke wrote: > >> Maybe we need a webkit-port-maintainers@ list that one could easily cc >> rather than trying to add people by hand? > > Sounds helpful. Not sure exactly how it would work, though. (How would you > add yourself to the list?) Subscribe through the listserv, just like webkit-dev? -- Dirk ___ webkit-dev mailing list webkit-dev@lists.webkit.org http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
Re: [webkit-dev] What is an "active port"? [WAS: Do you maintain OS(WINCE)?]
On Thu, Sep 15, 2011 at 12:26 PM, Dirk Pranke wrote: > On Thu, Sep 15, 2011 at 12:17 PM, Geoffrey Garen wrote: > > On Sep 14, 2011, at 1:02 PM, Dirk Pranke wrote: > > > >> Maybe we need a webkit-port-maintainers@ list that one could easily cc > >> rather than trying to add people by hand? > > > > Sounds helpful. Not sure exactly how it would work, though. (How would > you > > add yourself to the list?) > > Subscribe through the listserv, just like webkit-dev? > fwiw, I could totally see that working. Here's how I would use it: In general I would ignore the email except when I was on duty for keeping things green. Then, I would watch it carefully and get the right people to help out with the Chromium side of things as needed. dave > > -- Dirk > ___ > webkit-dev mailing list > webkit-dev@lists.webkit.org > http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev > ___ webkit-dev mailing list webkit-dev@lists.webkit.org http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
