Mailing-List: contact dev-help@jackrabbit.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@jackrabbit.apache.org
Date: Tue, 18 Feb 2014 23:57:28 +0000 (UTC)
From: "Tobias Bocanegra (JIRA)" <jira@apache.org>
To: dev@jackrabbit.apache.org
Message-ID: <JIRA.12695330.1392427549503.61599.1392767848399@arcas>
In-Reply-To: <JIRA.12695330.1392427549503@arcas>
References: <JIRA.12695330.1392427549503@arcas>
Subject: [jira] [Commented] (JCRVLT-31) Packages defined by a stream are
 copied to a file first
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


    [ https://issues.apache.org/jira/browse/JCRVLT-31?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13904834#comment-13904834 ] 

Tobias Bocanegra commented on JCRVLT-31:
----------------------------------------

note, that the NPE above only happens with JDK1.7

> Packages defined by a stream are copied to a file first
> -------------------------------------------------------
>
>                 Key: JCRVLT-31
>                 URL: https://issues.apache.org/jira/browse/JCRVLT-31
>             Project: Jackrabbit FileVault
>          Issue Type: Improvement
>            Reporter: Tobias Bocanegra
>            Assignee: Tobias Bocanegra
>             Fix For: 3.1
>
>
> in the current implementation the binary data of a package is always copied to a file before it's handled. because the meta-inf files might not be at the beginning of the stream. this might be a considerable performance drain when uploading large packages.
> in order to change this, the entire archive handling needs to be rewritten. i.e. the current way of copying the input stream to a tmp-file is "optimal" in this respect because the file is not regenerated when the package is accessed for installing.
> using a jar file a format might be possible - but is not controllable since we deliberately allow any kind of zips for packages that were not created by a jar-tool, package manager, vlt etc.
> however, another fast approach would be to scan-and-copy the package into the repository and extract at the same time the relevant metadata. since the stream is consumed during ValueFactory.createValue(),
> the proper packageId can be used to create the path and no tmp-location is needed.


--
This message was sent by Atlassian JIRA
(v6.1.5#6160)