OK here's the summary from quite an interesting irc session. Nobody likes this, but if we all think beyond it, we can release and eventually graduate. Moreover we can reduce our burden on the IPMC and our release managers who would have to answer for it. I think for these reasons there's more to gain working around the header policy than futilely fighting it.
So, here's the going advice: 1. play conservative to help avoid IPMC questions and also ease the job of our RM (Andrew B.) 2. this means that we'll add headers to everything in src with noted exceptions in test resources 3. any rat exception will be noted as an xml comment in the RAT config in the root pom.xml file Thanks for the participation in a colorful topic such as this :) -A On Thu, May 16, 2013 at 1:26 PM, Ignasi <[email protected]> wrote: > Just to make it clear, we'll be adding the license header in each > script fragment, so the jclouds "distribution" will hace the header > properly placed in each source file. > > What we will do is to remove the licenses at runtime, to be able to > compose those fragments in a readable way, but that will only happen > at runtime when generating a script to be uploaded to a node. > > On 16 May 2013 22:22, Andrew Bayer <[email protected]> wrote: > > Because each source file needs the license, not the binary. An Apache > > release is first and foremost a release of the source code, so if there > are > > any files in the source tree that *don't* have the license header, it'll > > raise questions. > > > > A. > > > > On Thu, May 16, 2013 at 1:21 PM, Adrian Cole <[email protected] > >wrote: > > > >> why would we need to justify lines of code inside scriptbuilder? that > >> collapses the header into one? Is there a rule that says we need > headers > >> randomly placed inside files templating systems generate? > >> > >> > >> On Thu, May 16, 2013 at 1:18 PM, Andrew Bayer <[email protected] > >> >wrote: > >> > >> > We'll have to justify that to the IPMC/mentors etc when we release - > be > >> > prepared for that. > >> > > >> > A. > >> > > >> > On Thu, May 16, 2013 at 1:15 PM, Adrian Cole <[email protected] > >> > >wrote: > >> > > >> > > Ignasi, > >> > > > >> > > I'm with you on this. Let's have a follow-up pull request to strip > >> > > comments inside scriptbuilder. > >> > > > >> > > -A > >> > > > >> > > > >> > > On Thu, May 16, 2013 at 1:14 PM, Ignasi <[email protected]> > >> > wrote: > >> > > > >> > > > I really don't like adding headers to script "fragments". I > >> understand > >> > > > the need to have a license header in a script, but when it comes > to > >> > > > compose them to produce one single file that will be copied to the > >> > > > deployed nodes and then executed, then we'll end up with scripts > with > >> > > > the license headers at random parts. > >> > > > > >> > > > Script execution would not be affected, but reading the script and > >> > > > seeing license texts here and there, IMHO gives a feeling of > >> something > >> > > > broken, or something that did not get properly generated. > >> > > > > >> > > > If we must add the license headers to the script fragments, I'd > like > >> > > > to consider modifying the scriptbuilder to strip comments. > >> > > > > >> > > > On 16 May 2013 22:11, Adrian Cole <[email protected]> > wrote: > >> > > > > cool. thanks > >> > > > > > >> > > > > > >> > > > > On Thu, May 16, 2013 at 1:09 PM, Andrew Bayer < > >> > [email protected] > >> > > > >wrote: > >> > > > > > >> > > > >> I changed a few places in src/test/resources to deal with > >> > > scriptbuilder > >> > > > >> functions that now spit out test headers - that's reasonable > to me > >> > > > being as > >> > > > >> far as we go. > >> > > > >> > >> > > > >> A. > >> > > > >> > >> > > > >> On Thu, May 16, 2013 at 1:04 PM, Adrian Cole < > >> > [email protected] > >> > > > >> >wrote: > >> > > > >> > >> > > > >> > Fair enough. I'll risk breaking folks to avoid perpetual > >> > > > explanations :) > >> > > > >> > > >> > > > >> > So, the creativity clause seems to help justify lack of > checking > >> > > > >> > src/test/resource, which only includes test expectation data. > >> I'm > >> > > > going > >> > > > >> to > >> > > > >> > make a call and continue to filter this out, as otherwise > we'd > >> > have > >> > > to > >> > > > >> > change our unit tests to emit license headers. > >> > > > >> > > >> > > > >> > I'll take care of the src/main/resources things (like script > >> > > > fragments) > >> > > > >> in > >> > > > >> > a separate commit as it will break unit tests. > >> > > > >> > > >> > > > >> > -A > >> > > > >> > > >> > > > >> > > >> > > > >> > On Thu, May 16, 2013 at 1:00 PM, Andrew Bayer < > >> > > [email protected] > >> > > > >> > >wrote: > >> > > > >> > > >> > > > >> > > And fwiw, I already added headers to a bunch of .sh files > to > >> > meet > >> > > > RAT > >> > > > >> > > checks. > >> > > > >> > > > >> > > > >> > > A. > >> > > > >> > > > >> > > > >> > > On Thu, May 16, 2013 at 12:46 PM, David Nalley < > [email protected] > >> > > >> > > > wrote: > >> > > > >> > > > >> > > > >> > > > On Thu, May 16, 2013 at 3:16 PM, Adrian Cole < > >> > > > >> [email protected]> > >> > > > >> > > > wrote: > >> > > > >> > > > > Hi, all. > >> > > > >> > > > > > >> > > > >> > > > > per > >> > > > >> > https://github.com/jclouds/jclouds/pull/6#issuecomment-18022286 > >> > > > >> > > > > > >> > > > >> > > > > jclouds includes a utility called scriptbuilder, which > >> > > generates > >> > > > >> > shell > >> > > > >> > > > > scripts from other fragments. We've not added license > >> > headers > >> > > > in > >> > > > >> the > >> > > > >> > > > past > >> > > > >> > > > > as these scripts are combined at runtime. > >> > > > >> > > > > > >> > > > >> > > > > Ex. you can imagine that doing a command like below, > the > >> > > > resulting > >> > > > >> > > shell > >> > > > >> > > > > script would senselessly have multiple ASF license > headers > >> > > > inlined. > >> > > > >> > > > > > >> > > > >> > > > > runScript = new StatementList(installJDK, addRoot); > >> > > > >> > > > > > >> > > > >> > > > > I seriously have objections about insisting adding > license > >> > > > headers > >> > > > >> to > >> > > > >> > > > > script fragments, not only from the efficiency concern, > >> but > >> > > also > >> > > > >> that > >> > > > >> > > it > >> > > > >> > > > > adds a chance of hard-to-troubleshoot bugs. For > example, > >> if > >> > > we > >> > > > >> added > >> > > > >> > > > > license headers to the script fragment for nohup, > >> everything > >> > > > that > >> > > > >> > uses > >> > > > >> > > > > nohup will have an extra 14 lines of comments, or we'd > >> have > >> > to > >> > > > >> write > >> > > > >> > > code > >> > > > >> > > > > to remove it. In cases where scriptBuilder is used as > EC2 > >> > > > instance > >> > > > >> > > data, > >> > > > >> > > > > it might push us over the limit. > >> > > > >> > > > > > >> > > > >> > > > > Bottom-line question is: > >> > > > >> > > > > > >> > > > >> > > > > Does the ASF require license header on inputs to > commands, > >> > > such > >> > > > as > >> > > > >> > > shell > >> > > > >> > > > > script fragments that are inputs to ScriptBuilder? > >> > > > >> > > > > > >> > > > >> > > > > -A > >> > > > >> > > > > >> > > > >> > > > So the default answer is that everything human-readable > >> > > requires a > >> > > > >> > > > license header. > >> > > > >> > > > There is an exception, namely: > >> > > > >> > > > > http://www.apache.org/legal/src-headers.html#faq-exceptions > >> > > > >> > > > > >> > > > >> > > > I went and looked at some of the functions and while I > might > >> > > agree > >> > > > >> > > > that something like abort.sh might qualify for the above > >> > > > exception, > >> > > > >> > > > something like setupPublicCurl.sh doesn't IMO. > >> > > > >> > > > > >> > > > >> > > > Additionally - you'll have folks (mentors and other IPMC > >> > > members) > >> > > > >> > > > reviewing this and their purpose is to catch problems - > so > >> you > >> > > (or > >> > > > >> the > >> > > > >> > > > release manager) will have to justify not including > licenses > >> > > > headers > >> > > > >> > > > for each of those license-header-excluded files. > >> > > > >> > > > > >> > > > >> > > > There has been a discussion on legal-discuss about > adding a > >> > > short > >> > > > >> form > >> > > > >> > > > license header for short files - that would be two > comment > >> > lines > >> > > > >> > > > instead of 16, but it is not established policy. Take a > look > >> > at > >> > > > that > >> > > > >> > > > thread and at links from that file. > >> > > > >> > > > http://markmail.org/thread/xvrxxkela4goxmk2 > >> > > > >> > > > > >> > > > >> > > > --David > >> > > > >> > > > > >> > > > >> > > > >> > > > >> > > >> > > > >> > >> > > > > >> > > > >> > > >> >
