| 1 |
On Tue, 2020-12-29 at 11:29 +0000, Peter Stuge wrote: |
| 2 |
> Michał Górny wrote: |
| 3 |
> > > I'm sure that there are numerous cases where libressl doesn't |
| 4 |
> > > work, |
| 5 |
> > > but that's no reason to dismiss cases where it *does*. |
| 6 |
> > |
| 7 |
> > Are you asking people to put an effort into maintaining something |
| 8 |
> > that |
| 9 |
> > can't be practically installed? |
| 10 |
> |
| 11 |
> No, I'm rather asking to change the level of commitment. |
| 12 |
> |
| 13 |
> I agree completely that it's unreasonable for Gentoo (worse, 1 |
| 14 |
> person!) |
| 15 |
> to continuosly patch the entire world for libressel. |
| 16 |
> |
| 17 |
> I'm asking to stop doing that, yet still enable the choice between |
| 18 |
> openssl and libressl where that is possible without patches, even |
| 19 |
> if that's only openntpd and one other package. |
| 20 |
> |
| 21 |
> The method for that choice could of course depend on the number of |
| 22 |
> packages where it applies. |
| 23 |
|
| 24 |
I'm pretty sure that soon enough one of Portage/Python dependencies is |
| 25 |
going to become a blocker for everything. |
| 26 |
|
| 27 |
> > |
| 28 |
> > > Did anyone gather actual numbers? |
| 29 |
> > |
| 30 |
> > $ find -name '*libressl*.patch' | grep -v dev-libs/libressl | wc -l |
| 31 |
> > 61 |
| 32 |
> |
| 33 |
> I'm more interested in the number of packages which can use either |
| 34 |
> library without patches (like openntpd?). This may be tricky to find |
| 35 |
> out. :\ |
| 36 |
|
| 37 |
It's never that simple. Roughly speaking, there are 3 levels |
| 38 |
of incompatibility with LibreSSL: |
| 39 |
|
| 40 |
1. Stuff that does not build against LibreSSL. |
| 41 |
|
| 42 |
2. Stuff that builds just fine but fails at runtime in unpredictable |
| 43 |
ways (e.g. Tor mentioned today). |
| 44 |
|
| 45 |
3. Stuff that builds and works 'fine' but ends up being crippled (e.g. |
| 46 |
doesn't support new algorithms). |
| 47 |
|
| 48 |
The first one is somewhat easy to test (e.g. via tinderbox). The other |
| 49 |
two are very hard. |
| 50 |
|
| 51 |
> > > > |
| 52 |
> > Actually, I see that someone has apparently forked libtls into |
| 53 |
> > libretls |
| 54 |
> > (the irony!) that works against OpenSSL [1]. |
| 55 |
> |
| 56 |
> To me, that proves value in the libtls API and in keeping libressl in |
| 57 |
> the tree in some capacity, even if not as an alternative to openssl. |
| 58 |
> |
| 59 |
> Maybe with a USE flag which makes it install only libtls (built from |
| 60 |
> libressl static libs), which could be one provider for a |
| 61 |
> virtual/libtls. |
| 62 |
|
| 63 |
I don't see why we would put an effort into that. I've just packaged |
| 64 |
dev-libs/libretls, and the maintenance effort in it seems really |
| 65 |
minimal. Trying to get the same result from libressl is just repeating |
| 66 |
the work. |
| 67 |
|
| 68 |
|
| 69 |
-- |
| 70 |
Best regards, |
| 71 |
Michał Górny |
Archives