On Mon, Aug 1, 2011 at 2:04 PM, Eike Rathke <[email protected]> wrote: > Hi Rob, > > On Saturday, 2011-07-30 17:41:28 -0400, Rob Weir wrote: > >> > Regarding Apache OpenOffice.org, LibreOffice, Symphony, RedOffice, >> > NeoOffice, BrOffice, EuroOffice, etc. as siblings that share similar >> > code in their genes from the ancestor OpenOffice.org, this approach has >> > a shortcoming in collaboration. >> >> The purpose of ooo-security is security. Any other concerns, >> including collaboration with other projects, or even including >> collaboration within this project, are secondary. > > I was referring collaboration on security, for the benefit of all. >
Committers chosen by the PPMC collaborate on resolution of security issues today on the ooo-security list. If someone wants to take part in this collaboration, then they are welcome to do it. It starts with becoming a committer on the project. >> > Given that usually users / developers / security experts report >> > a problem to _their_ project's security team, this setup would mean >> > a one-way communication if that project decided to inform AOOo, actually >> > being a pre-notification in that direction. What is that project >> > expected to do from that point on? Wait until AOOo publishes the >> > disclosure and fix? I think the project would develop its own fix and, >> > hopefully, share it with AOOo (which would involve a CLA or a permissive >> > license or a grant) and also with other siblings. >> > >> >> We've already discussed and generally agreed that we can pre-notify >> other related projects about vulnerabilities and fixes. But >> pre-notification is not the same as saying that the other projects >> must be signed up to on ooo-security. > > Well, my example was about pre-notification in the other direction, from > siblings to AOOo. > Sorry, I misunderstood. I assume any other project would follow their own rules. Ideally, there would be a pre-notification state that stands between that team's initial investigation and proposed resolution and public notification. That is what some other Apache projects are doing, and that is what I'm proposing for us. >> > Now if siblings develop fixes independently because AOOo security runs >> > as a strict Apache closed coterie, we may get into the situation where >> > fixes are developed in parallel, maybe with different solutions or even >> > contradictory. I think the best would be if efforts would be bundled >> > instead and the best of all possible solutions shared as >> > pre-notification with siblings. >> > >> >> There is nothing wrong with pre-notification. That was my step #3 above. >> >> Your example seems to be assuming that a single reporter reports the >> same flaw to OpenOffice, LibreOffice, Symphony, etc., independently, > > No, my example was assuming that a single reporter reports only to one > project, which I think is the usual case. Further assumption was that > project would decide to pre-notify AOOo, knowing the same code is > affected. > OK. That's all true. To the extent that we're not working off a single code base, the possibility that "fixes are developed in parallel, maybe with different solutions or even contradictory" exists everywhere. This is not something special with security. It will occur with every are of the products. But with pre-notification, this doesn't need to be a problem. If we pre-notify with a patch under Apache 2.0, then anyone can use that patch. If others do the same, then we could as well. >> and that these projects develop patches independently. I suspect that >> this occurrence is quite rare. But I don't see that harm if it does >> happen. The fact is these projects are not sharing code today, >> period. > > Au contraire, you're right the projects aren't sharing code, in that > case a fix and update would be sufficient, but most projects still use > the same if not identical code in areas where security issues may be > discovered. > I don't think we're in disagreement here, and if we shared code we'd have less overlapping work. I'm just pointing out that security vulnerability patches amount to something like 0.1% of all code changes that we could be sharing. So if you are looking for meaningful ways to collaborate with LibreOffice then this is the wrong place to look. I'd focus on parts where we can work together openly, like in file import/export filers, plugins, etc. >> This is not just a statement about security. If we want to >> improve that situation with security specifically then we have two >> easy ways forward: >> >> 1) Develop the patches at Apache under a license that allows all other >> products to use it >> >> 2) Have security experts from LibreOffice developers join as Apache >> committers so they can then join the ooo-security list and contribute >> directly there. > > An ideal situation I support. And btw, I wasn't talking of only > LibreOffice.. > Good. > >> > The problem here seems to be the perceived requirement that the Apache >> > governance would allow only PMC members on a project's security list. >> > However, I didn't see that requirement, or it's not available somewhere >> > under http://apache.org/security/ >> > >> > Additionally http://apache.org/security/committers.html states that >> > >> > | Information may be shared with domain experts (eg colleagues at your >> > | employer) at the discretion of the project's security team providing >> > | that it is made clear that the information is not for public disclosure >> > | and that [email protected] or the project's security mailing list must >> > | be copied on any communication regarding the vulnerability. >> > >> > This IMHO allows also to have selected members of sibling projects as >> > domain experts on the security list, if I interpret "at the discretion >> > of the project's security team" correctly. >> > >> >> There is nothing there that talks about having these domain experts be >> list members. In fact, it specifically talks about "copying" >> [email protected]. That would not be necessary if the domain >> experts were security list members, since all list project security >> traffic is copied automatically to security.apache.org as well. Thus >> the information sharing implied here is through means other than >> subscription to the list. > > To me it only sounds like that for any communication regarding the > vulnerability the project's (or apache) security list must be copied, so > information is shared when the domain expert communicates with a list > member and even if he consults another colleague. Of course the language > implies that the domain expert is not a list member, but "at the > discretion of the security team" is a degree of freedom to decide whom > to copy when, and if the team decides to get another expert on board on > a regular basis that's just fine. > > Also, if all traffic were automatically >> copied to 3rd party domain experts via their subscription to the list, >> then there would be zero discretion being exercised. > > Given that only a hand full of known, respected and trusted persons > would be invited I don't see much difference whether they are Apache > committers or not, provided that the patches developed will be > contributed under AL2. > Remember, on any Apache list, we can bring in non-subscribers for a discussion. We just send an email to the list and cc the 3rd party. If the 3rd party then does a "reply to all", their email will go through to the list. Since the list is intended for 3rd party (public) submission of reports, anyone can post to it. So we can have a conversation on a specific issue on the ooo-security list, and include 3rd parties. I have no problems with this. What I oppose is having 3rd parties subscribe to the list so they automatically and by default get every single one of the raw, unverified, unconfirmed, un-anonymized private reports that are submitted to the project from the public. If there are 3rd parties who are respected and trusted and committed to this project, and knowledgeable about OOo security, then I invite them to state so on this list and I'll gladly propose them as committers. -Rob > Eike > > -- > PGP/OpenPGP/GnuPG encrypted mail preferred in all private communication. > Key ID: 0x293C05FD - 997A 4C60 CE41 0149 0DB3 9E96 2F1A D073 293C 05FD >
