On Mon, Nov 16, 2009 at 2:15 PM, Paul Lindner <[email protected]> wrote:
> Agreed on all points. I also think that with some tweaks this patch will > be > a great addition -- especially when combined with the hash checking in > JsServlet. > > John, can you elaborate on versioned iframe URLs? Don't you need to attach > a security token to iframe URLs to insure that data pipelining etal work > properly, which invalidates any caching benefit? > Yes - this is the use case that reduces the value of feature externalizing "somewhat". YMWV. Of course, the token isn't needed on the query for client-processed preloads/pipelining, or gadgets w/o these features. > On Mon, Nov 16, 2009 at 2:11 PM, John Hjelmstad <[email protected]> wrote: > > > Agreed that the &libs parameter should IMO definitely be taken into > > account, > > as it allows optimization across multiple simultaneous gadget renders > (and > > offers control to the container as to JS optimization). > > > > Consider gadget G1 includes features (A, B, C), G2 includes (B, C, D), G3 > > includes (C, D, E) rendered at once. &libs=A:B:C:D:E can be included in > > each > > for optimal JS loading (esp. in the cold-cache case). > > > > As it is, this patch disables &libs wholesale on a true value. > > > > It might also be worth mentioning that with versioned IFRAME URLs, the > > value > > of externalizing inlined code reduces somewhat as well. Any > analysis/stats > > on the impact of implementing this change as it stands? > > > > On Mon, Nov 16, 2009 at 1:55 PM, Paul Lindner (JIRA) <[email protected] > > >wrote: > > > > > > > > [ > > > > > > https://issues.apache.org/jira/browse/SHINDIG-1227?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12778576#action_12778576 > > ] > > > > > > Paul Lindner commented on SHINDIG-1227: > > > --------------------------------------- > > > > > > I wonder if it would be useful to generate two script tags? > > > > > > One for the externed libs from the libs= param and one for the > remainder. > > > > > > forced libs would then be a common set of libraries that all apps share > > and > > > the rest would be cached per app. > > > > > > > > > > Allow JavaScript features required by a gadget to be externalized on > > > demand > > > > > > > > > > --------------------------------------------------------------------------- > > > > > > > > Key: SHINDIG-1227 > > > > URL: > > https://issues.apache.org/jira/browse/SHINDIG-1227 > > > > Project: Shindig > > > > Issue Type: New Feature > > > > Components: Java > > > > Reporter: chirag shah > > > > Attachments: SHINDIG-1227.patch.diff > > > > > > > > > > > > This patch introduces a configuration parameter in the properties > file > > > "shindig.gadget-rewrite.externalize-feature-libs" > > > > When this parameter is set to true, RenderingGadgetRewriter will > gather > > > all features requested by the gadget and simply add them to the set of > > > extern libraries. As a result, the browser will see <script src=' > > > http://shindig-host/gadgets/js/feature1:feature:2:feature3'><script> > and > > > get the features specified in the path. > > > > When this parameter is set to false (the default), > > > RenderingGadgetRewriter will do what it normally does. Inline features > > > requested by the gadget. > > > > The benefit of setting externalize-feature-libs to true is that > > browsers > > > can cache features requested by the gadget. > > > > > > -- > > > This message is automatically generated by JIRA. > > > - > > > You can reply to this email to add a comment to the issue online. > > > > > > > > >
