Where to find recipes?
|
|
| 1 2 |
Tarek Ziade
|
2008/3/12, Raphael Ritz <[hidden email]>: Dylan Jay wrote: Yes Dylan, I think the idea is great and is not a problem with the other PSC works I am doing at the moment. If you dig the idea and come up with some kind of "protocol" description for the client/sever exchanges, and an initial stats view idea, we can probably come up with something quite quickly. And maybe think of a pluggable way to add this kind of feature to a PSC instance. (I can help on the coding after this initial step) Regards Tarek Raphael ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ Plone-developers mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/plone-developers |
||||||||||||||||
Dylan Jay-3
|
Tarek Ziade wrote:
> > > 2008/3/12, Raphael Ritz > <[hidden email] > <mailto:[hidden email]>>: > > Dylan Jay wrote: > > Tarek Ziadé wrote: > > [..] > > > >> I think this could be a feature in PloneSoftwareCenter, as > anyone setting up > >> a software center would like to have such feature I guess. > >> > >> What I would do is create a recipe prototype > (plone.softwarecenter.stats?) > >> and see what would be the required web services on server-side > (XML-RPC?) > >> to get the stats. > >> > >> Then we could see how an add-on in PSC could be integrated to > provide the > >> webservices, > >> to collect the data and provide a statistical view > >> > >> What others think ? (I think some strategic summit tasks are > dealing with > >> this too) > > > > I think its an idea that could work well but I think the > plone.org <http://plone.org> > > software center stuff is undergoing changes at the moment so > maybe its > > not the time. > > > ??? You are aware that Tarek is one of the major drivers of these > changes, are you? So now is a good time, I'd say. > > > Yes Dylan, I think the idea is great and is not a problem with the other > PSC works I am > doing at the moment. > > If you dig the idea and come up with some kind of "protocol" description > for the client/sever exchanges, > and an initial stats view idea, we can probably come up with something > quite quickly. And maybe think > of a pluggable way to add this kind of feature to a PSC instance. (I can > help on the coding after this initial > step) Ok, I'll give this some thought. Initially my modelling is something like this Website *- Buildout *-* Release *- Software - we can link to Software and/or Release to a buildout count which would work as a kind of vote tally. It would be good to sort everything by this number automatically. - Software view could then list "sites which use this software". - A site view can have links to "software used in this site" as well as be a content item owned by the site owner, giving a description of the site. - A Software/release page has a section "other software used with this". - A listing of all websites that use plone I'm not sure the best way of tying this all togeather. It could be references, plone.relations, a sql db or perhaps something else? Anyone have any ideas? Also would porlets, viewlets or some other methods be used to create the links from the software center views to website or buildout views? Additionally on the buildout side we'd need a unique id for a buildout such that on each production build the data is uploaded to the right place on plone.org. My model above assumes there is only ever a single buildout per website. We could then use the website url to lookup which buildout info to overwrite. Is this a safe assumption? Would we need to protect against cheating by people creating lots of fake urls? On the buildout side I imagine it to look something like [plonestats] websites = http://www.plone.org eggs = ${instance:eggs} products = ${instance:products} user = djay password = .... Is there a more clever way of doing this via introspecting the buildout? so the buildout author does less work. Once uploaded is it easy enough to match the eggs and products to items in the software center? or is it more messy than that? ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ Plone-developers mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/plone-developers |
||||||||||||||||
|
Dylan Jay-3
|
In reply to this post
by Tarek Ziade
Tarek Ziade wrote:
> > > 2008/3/12, Raphael Ritz > <[hidden email] > <mailto:[hidden email]>>: > > Dylan Jay wrote: > > Tarek Ziadé wrote: > > [..] > > > >> I think this could be a feature in PloneSoftwareCenter, as > anyone setting up > >> a software center would like to have such feature I guess. > >> > >> What I would do is create a recipe prototype > (plone.softwarecenter.stats?) > >> and see what would be the required web services on server-side > (XML-RPC?) > >> to get the stats. > >> > >> Then we could see how an add-on in PSC could be integrated to > provide the > >> webservices, > >> to collect the data and provide a statistical view > >> > >> What others think ? (I think some strategic summit tasks are > dealing with > >> this too) > > > > I think its an idea that could work well but I think the > plone.org <http://plone.org> > > software center stuff is undergoing changes at the moment so > maybe its > > not the time. > > > ??? You are aware that Tarek is one of the major drivers of these > changes, are you? So now is a good time, I'd say. > > > Yes Dylan, I think the idea is great and is not a problem with the other > PSC works I am > doing at the moment. > > If you dig the idea and come up with some kind of "protocol" description > for the client/sever exchanges, > and an initial stats view idea, we can probably come up with something > quite quickly. And maybe think > of a pluggable way to add this kind of feature to a PSC instance. (I can > help on the coding after this initial > step) Ok, I'll give this some thought. Initially my modelling is something like this Website *- Buildout *-* Release *- Software - we can link to Software and/or Release to a buildout count which would work as a kind of vote tally. It would be good to sort everything by this number automatically. - Software view could then list "sites which use this software". - A site view can have links to "software used in this site" as well as be a content item owned by the site owner, giving a description of the site. - A Software/release page has a section "other software used with this". - A listing of all websites that use plone I'm not sure the best way of tying this all togeather. It could be references, plone.relations, a sql db or perhaps something else? Anyone have any ideas? Also would porlets, viewlets or some other methods be used to create the links from the software center views to website or buildout views? Additionally on the buildout side we'd need a unique id for a buildout such that on each production build the data is uploaded to the right place on plone.org. My model above assumes there is only ever a single buildout per website. We could then use the website url to lookup which buildout info to overwrite. Is this a safe assumption? Would we need to protect against cheating by people creating lots of fake urls? On the buildout side I imagine it to look something like [plonestats] websites = http://www.plone.org eggs = ${instance:eggs} products = ${instance:products} user = djay password = .... Is there a more clever way of doing this via introspecting the buildout? so the buildout author does less work. Once uploaded is it easy enough to match the eggs and products to items in the software center? or is it more messy than that? ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ Plone-developers mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/plone-developers |
||||||||||||||||
Tarek Ziadé
|
In reply to this post
by Dylan Jay-3
Cool ! Sounds good. I think we could also have a big number on the top right corner of the Software view, à la digg, that indicates how many web sites uses it. For the list, I think we should rather have a separate page to avoid long lists for packages like plone.recipe.plone for instance. Maybe we could also deal differently recipe package and other packages. From a PSC point of view, I think we could deal it with a btree that holds website objects. that would make it easier to hook things together. Now having this feature as a separate package sounds good too, (like a plugin things) so I am not sure yet of what would be the best way. The problem for the sql db is that it adds a third-party server to install together with the plone instance, which is fine for plone.org but might be painful for people that wishes to run PSC out of the box, so if we do so, we need to make sure the default storage is ZODB. Here's what we can do: the user registers a website, let's say "http://plone.org". We generate a new id, ad send him by mail a hash of the new web site id, and ask him to put it in a custom metadata in the headers of the front page of the site. Then to validate it has be done. This would let us control that he owns the site, then keep a unique key. That means we would need a managment screen where a user gets a list of registered websites. I think we can use zc.buildout APIs here to do some introspection, and get infos on what is being used. For the user/password, we need to keep it out of the buildout, to avoid having them released or available in code repositories lovely.buildouthttp can be used there as an extension as long as the server side raise the proper authentication error. So the plonestats section would be dead simple: [plonestats] recipe=plone.softwarecenter.stats websites = http://www.plone.org It shouldn't get too complicated anyway, if we want to have a lot of people using it. In the future, if all this gets done, I can think of a question in the UnifiedInstaller "Do you whish to send stats to plone.org ?" Or something like that. But that's another story... Now for the stats that are being retrieved, we need of course to drop the ones that are not listed in PSC. Another feature we might need to add is an option at the Package level in PSC for people that does not wish to be linked to the stats maybe. Tarek |
||||||||||||||||
| Free Embeddable Forum Powered by Nabble | Help |
