I am the OP.
For what it is worth, the current discussion is actually precisely what I was aiming at. I agree with your analysis of my Intended goals but completely disagree that SHA1 alone is any sort of guarantee.. To be precise, I don't just think that it doesn't provide much, but rather that alone it provides none at all. This is because I have no idea who produced a given SHA1 - whether it was Bastien, or a MITM attacker, or just someone who compromised the server.
Of course I don't care much about whether the code would be available via gitolite or gogs, or at least much less than that something would be available. Together with signing of releases tags it would be magnificent and actually *would* provide a significant level of guarantee (at least for me).
Finally, I'd like to add that if the decision regarding the git hosting service comes down to such consideration, that gogs seems much more polished from a user perspective.
I think that the original question was looking at a different problem,
and discussion of hosted tooling may be a distraction. The issues that
normally come up for cyber-security discussions of distribution need to
be looked at. The following is a start at organizing those for
org-mode.
I think that the problem that started this discussion was a question
about whether an org-mode user could verify the integrity and
authenticity of an org-mode version distribution, or perhaps the
integrity and authenticity of an installed version. It may be in the
context of ELPA for that user. A full answer for all org-mode users
would certainly include ELPA.
I personally use a combination of "git clone" and "git checkout" to
manage my org-mode versions. That provides adequate protection at the
moment, but SHA1 is not sufficient in the long term.
R Horn
address@hidden
**** [2016-07-03 Sun 12:15] Org security
***** Org cyber-security
****** Process questions
******* How are security flaws discovered
******** Is there a known public reporting channel?
******** Are they privately and publicly tracked.
For example, are CVE numbers obtained for public tracking, reporting, status, etc.
******** Are patches tracked relative to CVE numbers?
******** Are security-only patches and releases made?
- For current development and release versions?
- For widely deployed older versions?
******* Are security processes understood, maintained, staffed, performed?
****** Development process related questions
I'll let these wait for now.
****** Release and Distribution process related questions
******* Org has multiple release and distribution chains.
- The underlying git structure is used by developers and some of the end users
- Direct git clone is used by some, and is one distribution chain
- Git hosting is used by some. This is git plus extra hosting facilities
- ELPA is used by some.
- Tar-ball distribution is still used by some
- Bundled distributions (Red Hat, Ubuntu, etc.) are used by some
- Other distribution methods can be ignored for now (in my opinion)
******* How are releases identified?
******** GIT
- GIT identifies changes and tagged releases by SHA1 hash. SHA1 hash remains excellent as integrity verification against accidental damage. SHA1 is end of life against intentional attack. It is expected to be vulnerable to well funded attackers by 2020.
- see also below on verification and identification
******** GIT hosting
depends on the host.
******** ELPA
ELPA is seriously lacking, at least in terms of documentation. This is probably the source of the initial user question. ELPA can identify version strings.
******** Tar-ball
Nothing appears to be set up for robust identification of tar balls for org-mode.
******** Bundled distributions
Distributions have identification systems that do not always match the org-mode system.
******* How does distribution process protect integrity and authentication of a release?
- Is there integrity check and authenticated verification from development release to end user installation?
- Can an end user verify the integrity and authenticate the installed software?
******** GIT
- GIT can provide PGP signing for tagged releases. Org does not currently use this feature. PGP signing is considered secure for integrity and authentication verification, provided the security processes are sufficient in the development release system.
- GIT can provide PGP signing for individual updates. Org does not currently use this feature. PGP signing at the individual update level is often too burdensome for projects. Every committer must be familiar with PGP signing and properly protect their keys. There is a key management headache for project administration.
******** Git Hosting
- Depends of the hosting. Hosted verification often changes the risks rather than eliminate them. If hosting facility protection is used instead of git PGP signing, the risk changes from PGP management flaws into host service user management flaws. Instead of PGP headaches there are the headaches from lost SSH keys and user spoofing.
******** ELPA
- ELPA does not appear to have any verification or authentication capability
******** Tar-ball
Nothing appears to be set up for verification or authentication of tar-balls for org-mode
******** Bundled distributions
The bundled distributions use signed packages. Some distributions can verified installed packages, while others don't provide this feature. The distribution chain from "bundler" to end user install is well protected. Protection from org-mode developer to "bundler" is unknown.
******* Update process
- When a new release is made, is integrity and authentication verified? to distributor? to end user? How difficult is this?
- Is there a fast path for security patches?
- Is there a special notification path to users for security patches?
- Are security updates for older versions easily obtained and installed?