gistfile1.txt

[04/12/11 10:32] <sebersole> hardy: anyway, back to that Metadata discussion
[04/12/11 10:32] <sebersole> this is what we need to decide
[04/12/11 10:32] <ge0ffrey> sorry for interrupting :) tnx for the info
[04/12/11 10:32] <stliu> ge0ffrey, good, let me know if you need me do something
[04/12/11 10:32] <hardy> one sec. let me just leave the JBoss talk
[04/12/11 10:32] <sebersole> there could be *something* above Metadata
[04/12/11 10:32] <hardy> then you get my undevided attention
[04/12/11 10:32] <sebersole> hardy: ah, np
[04/12/11 10:33] <ge0ffrey> oh yea, one more thing: Can I release 2.3.5 of the maven-jdocbook-plugin?
[04/12/11 10:33] <sebersole> ge0ffrey: you maintain it now, do what you wish ;)
[04/12/11 10:34] <hardy> sebersole: now :-)
[04/12/11 10:34] <ge0ffrey> k :) I 'll do the job they the aliens eat me or someone better comes along
[04/12/11 10:34] <sebersole> hardy: wow that was a long call
[04/12/11 10:34] <hardy> I remember now as well that I was wondering how this in the end bootstraps
[04/12/11 10:34] <hardy> 1 1/2 hours
[04/12/11 10:34] <stliu> 90' call
[04/12/11 10:34] <sebersole> too bad i missed it :D
[04/12/11 10:34] <sebersole> shucks
[04/12/11 10:35] <hardy> some parts were interesting, others ...
[04/12/11 10:35] <hardy> when you talk about users and Metadata you are talking about how the users adds annotated classes or xml files to it, right?
[04/12/11 10:36] <sebersole> hardy: so either (1) Metadata is the exposed API (2) Metadata gets split up into api/spi/internal (3) we have another construct with whcih user interacts
[04/12/11 10:36] <sebersole> hardy: yes
[04/12/11 10:36] <stliu> smarlow_call, i thought u're going to demo jpa, isn't it?
[04/12/11 10:37] <hardy> from that perspective MetaData is exposinfg really to much
[04/12/11 10:37] <sebersole> hardy: sure
[04/12/11 10:37] <sebersole> hence (2)
[04/12/11 10:37] <hardy> right
[04/12/11 10:37] <hardy> how does MetadataSourceQueue fit in here?
[04/12/11 10:38] <sebersole> well i think thats no longer needed
[04/12/11 10:38] <smarlow_call> stliu:  first I heard about that :)
[04/12/11 10:38] <hardy> this is just a container we use to store the different configuration sources
[04/12/11 10:38] <sebersole> right
[04/12/11 10:38] <sebersole> its based on the old code
[04/12/11 10:38] <sebersole> so its where we keep stuff after its been added
[04/12/11 10:38] <sebersole> until we process it
[04/12/11 10:38] <sebersole> i still want to keep that general flow
[04/12/11 10:38] <hardy> so we are looking for an API which allows the user to add classes, packages, xml files which then get added to MetadataSourceQueue
[04/12/11 10:38] <sebersole> so you have 2 phases
[04/12/11 10:39] <hardy> :-)
[04/12/11 10:39] <sebersole> hardy: well like i said i think MetadataSourceQueue may no longer be needed
[04/12/11 10:40] <sebersole> i mean the other option is to yeah expose it to the users
[04/12/11 10:40] <sebersole> as the first "phase" (adding sources)
[04/12/11 10:40] <sebersole> the second phase being the processing of all those sources in prep for building a session factory
[04/12/11 10:41] <sebersole> so the "api" part here is really the addXXX stuff from Configuration
[04/12/11 10:41] <hardy> right
[04/12/11 10:42] <sebersole> though in my opinion we want to do a better job of distinguishing xml type up front
[04/12/11 10:43] <hardy> what do you mean w/ that?
[04/12/11 10:43] <sebersole> right now we essentially have to read the XML source once to see what type it is so we know what validation to apply
[04/12/11 10:43] <sebersole> it would be better if the user told us what type it is
[04/12/11 10:43] <sebersole> addResource(...) for example
[04/12/11 10:43] <hardy> so you want different addXml... methods
[04/12/11 10:43] <ge0ffrey> sebersole: looks like I don't have commit rights on the jdocbook plugin. My username is ge0ffrey (with a zero) and the svn url is https://svn.jboss.org/repos/maven/plugins/jdocbook/maven-jdocbook-plugin/trunk
[04/12/11 10:43] <hardy> or an addXml(filename, type)
[04/12/11 10:43] <sebersole> hardy: yeah, or
[04/12/11 10:44] <sebersole> addXml is inherently hbm.xml
[04/12/11 10:44] <sebersole> and we add:
[04/12/11 10:44] <sebersole> addOrmXml
[04/12/11 10:44] <sebersole> addOrmXml( source, version )
[04/12/11 10:44] <hardy> I think having this distinction is a good idea
[04/12/11 10:45] <hardy> the only drawback is that  we put more work onto the user
[04/12/11 10:45] <sebersole> not sure how that plays with persistence.xml
[04/12/11 10:45] <sebersole> yes and no
[04/12/11 10:45] <sebersole> i mean surely they know what types their files are
[04/12/11 10:46] <sebersole> hardy: actually this is really much less of an issue now iiuc because hbm.xml will be xsd
[04/12/11 10:46] <sebersole> we could really (1) read the xml and peek at the namespace, etc
[04/12/11 10:47] <sebersole> (2) apply validation directly based on that
[04/12/11 10:48] <hardy> so is this now an argument pro having a single addXml method?
[04/12/11 10:48] <-- emmanuel has left this server (Ping timeout: 248 seconds).
[04/12/11 10:49] <sebersole> hardy: well i am not sure how that plays with stuff like default values
[04/12/11 10:50] <-- smarlow_call has left this server (Ping timeout: 246 seconds).
[04/12/11 10:50] <hardy> which default values?
[04/12/11 10:52] <-- maxandersen has left this server (Quit: Leaving.).
[04/12/11 10:52] <sebersole> hardy: the ones from the DTD
[04/12/11 10:55] <hardy> not sure. you lost me on this one. going back to MetaData. I think it should not be exposed to the user. Instead we should expose an interface (maybe MetadataSource) with just the different addXYZ methods. Including splitting between addXml and addOrmXml
[04/12/11 10:57] <sebersole> hardy: yes, it should be "scaled down" whatever it is, but...
[04/12/11 10:57] <sebersole> it cannot be just an interface
[04/12/11 10:57] <sebersole> this goes back to the discussion yesterday
[04/12/11 10:58] <sebersole> the decision (short term) was for the user to instantiate this thing
[04/12/11 10:58] <hardy> right, so it is a class which the user can instantiate and populate
[04/12/11 10:59] <hardy> so that would be really alternative (3)
[04/12/11 10:59] <hardy> referring to
[04/12/11 10:59] <hardy> so either (1) Metadata is the exposed API (2) Metadata gets split up into api/spi/internal (3) we have another construct with whcih user interacts
[04/12/11 11:00] <hardy> do you see a benefit in (2)?
[04/12/11 11:00] <sebersole> hardy: longer term perhaps when if this becomes a service
[04/12/11 11:00] <sebersole> the api is the service contract
[04/12/11 11:01] <sebersole> the spi is the stuff the delegates know about
[04/12/11 11:01] <sebersole> and the impl is the impl :)
[04/12/11 11:02] <hardy> works for me as well. the most important is -1 to (1)
[04/12/11 11:08] <sebersole> hardy: ok, i'l start working on that
[04/12/11 11:08] <hardy> ok
[04/12/11 11:08] <sebersole> do you think it makes sense to keep around MetadataSourceQueue?
[04/12/11 11:09] <sebersole> or just subsume that stuff into Metadata?
[04/12/11 11:09] <sebersole> some of that is strictly HBM too
[04/12/11 11:10] <hardy> I think it becomes obsolete
[04/12/11 11:10] <sebersole> extendsqueue too possibly
[04/12/11 11:10] <hardy> or in case of option (3) i saw MetadataSourceQueue becoming the MetadataSource I was talking about
[04/12/11 11:11] <hardy> right, ExtendsQueue as well
[04/12/11 11:12] <hardy> but it is hbm specific anyways
[04/12/11 11:12] <hardy> what I want to say w/ that is that I don't really know how important this class is for hbm mapping ;-)
[04/12/11 11:13] <hardy> btw, do we just keep pushing changes to master?
[04/12/11 11:19] --> gbadner has joined this channel ([email protected]).
[04/12/11 11:19] <sebersole> hardy: yes
[04/12/11 11:19] <sebersole> thats the general idea
[04/12/11 11:19] <hardy> ok
[04/12/11 11:19] <sebersole> do little pieces of work, integrate
[04/12/11 11:20] <sebersole> hardy: here is my concern with (3)
[04/12/11 11:20] <gbadner> sebersole, pong
[04/12/11 11:20] <sebersole> gbadner: sec
[04/12/11 11:20] <sebersole> although this discussion with hardy involves you too
[04/12/11 11:21] <sebersole> and the questions i asked last night
[04/12/11 11:21] <sebersole> hardy: here is my concern with (3)
[04/12/11 11:21] <gbadner> k, can you post your conv on pastebin?
[04/12/11 11:21] <sebersole> so we define a "MetadataSource"
[04/12/11 11:21] <sebersole> and users use that
[04/12/11 11:21] <sebersole> how do we get from that to Metadata ?
[04/12/11 11:22] <sebersole> MetadataSource will need access to services too btw
[04/12/11 11:22] <sebersole> ClassLoaderService
[04/12/11 11:22] <sebersole> to properly resolve resource lookups