http spews many "private" events before any real data events. because layout can occupy the UI thread for relatively long periods of time, processing the top level document, this can result in large delays before actually downloading any sub documents such as inline images. by pushing more of http onto the socket thread, we can ensure that the first plevent posted via http corresponds to real data for http's listener.

targetting 0.9.1

this work is happening on the branch: HttpConnectionMagic_20010415_BRANCH and currently includes: netwerk/protocol/http, netwerk/build, and netwerk/test

actually, it was necessary to branch all of netwerk/

the following files are also on the branch: dom/src/base/nsGlobalWindow.cpp content/html/document/src/nsHTMLContentSink.cpp content/html/document/src/nsHTMLDocument.cpp content/xml/document/src/nsXMLDocument.cpp content/xml/document/src/nsXMLDocument.h content/xbl/src/nsXBLService.cpp content/build/nsContentHTTPStartup.cpp content/build/gbdate.pl layout/html/forms/src/nsFormFrame.cpp xpfe/bootstrap/nsAppRunner.cpp xpfe/browser/src/nsBrowserInstance.cpp xpfe/components/bookmarks/src/nsBookmarksService.cpp xpfe/components/search/src/nsInternetSearchService.cpp xpfe/components/xfer/src/nsStreamTransfer.cpp xpfe/components/ucth/src/nsUnknownContentTypeHandler.cpp docshell/base/nsDocShell.cpp docshell/base/nsWebShell.cpp gfx/src/nsImageNetContextAsync.cpp uriloader/base/nsDocLoader.h uriloader/base/nsDocLoader.cpp uriloader/base/nsURILoader.cpp uriloader/exthandler/nsExternalHelperAppService.cpp modules/plugin/nglsrc/nsPluginHostImpl.cpp intl/chardet/src/nsObserverBase.cpp embedding/browser/webBrowser/nsWebBrowserPersist.cpp extensions/cookie/nsCookieHTTPNotify.h extensions/cookie/nsCookieHTTPNotify.cpp mailnews/compose/src/nsMsgCompUtils.cpp mailnews/compose/src/nsURLFetcher.cpp mailnews/compose/src/nsURLFetcher.h extensions/xmlextras/base/src/nsXMLHttpRequest.h extensions/xmlextras/base/src/nsXMLHttpRequest.cpp extensions/xmlextras/base/public/nsIXMLHttpRequest.idl mailnews/absync/src/nsAbSyncPostEngine.cpp

the branch has landed.

qa to me. Anything I can do to verify this?

prolly not.. jrgm confirmed the 10% speed up.. so that should be sufficient.

Yeah, indeed, I hath seen this effect, so vrfy.