| 1 |
On Mon, 2020-12-28 at 23:18 +0000, Peter Stuge wrote: |
| 2 |
> Michał Górny wrote: |
| 3 |
> > > A. It is a distinct implementation with probably /quite some/ |
| 4 |
> > > stable |
| 5 |
> > > compatibility, meaning that it will work perfectly fine as an |
| 6 |
> > > alternative in many cases. |
| 7 |
> > |
| 8 |
> > Except that it doesn't, as has been proven numerous times. |
| 9 |
> |
| 10 |
> I'm sure that there are numerous cases where libressl doesn't work, |
| 11 |
> but that's no reason to dismiss cases where it *does*. |
| 12 |
|
| 13 |
Are you asking people to put an effort into maintaining something that |
| 14 |
can't be practically installed? 'I don't use a single Qt application |
| 15 |
(yet!), so I demand you support LibreSSL for me!' |
| 16 |
|
| 17 |
A side effect of this approach is that users will be tricked into using |
| 18 |
LibreSSL, only to discover that they eventually have to transition |
| 19 |
their systems back. We have been there with LibAV. However, unlike |
| 20 |
LibAV, transition between SSL providers is non-trivial. |
| 21 |
|
| 22 |
> Did anyone gather actual numbers? |
| 23 |
|
| 24 |
$ find -name '*libressl*.patch' | grep -v dev-libs/libressl | wc -l |
| 25 |
61 |
| 26 |
|
| 27 |
That's just the tip of the iceberg. Nobody's going to be able to even |
| 28 |
guess how many upstream have accepted *bad* patches. How many packages |
| 29 |
are forcing obsolete algorithms because someone added '!libressl' |
| 30 |
conditions because LibreSSL keeps pretending to be a newer OpenSSL |
| 31 |
version without actually implementing its API. |
| 32 |
|
| 33 |
> > > B. It brings its own TLS API, a unique feature which by itself |
| 34 |
> > > warrants the package. |
| 35 |
> > |
| 36 |
> > ...which by itself has no future |
| 37 |
> |
| 38 |
> That's arrogant and silly coming from anywhere but upstream. |
| 39 |
> |
| 40 |
> You can argue that you will never use the API in your TLS programs, |
| 41 |
> but even then that says really nothing about the API provider itself. |
| 42 |
|
| 43 |
That's my opinion and I have a right to have it without being insulted. |
| 44 |
|
| 45 |
I don't dispute whether it's good. I dispute whether tightly binding |
| 46 |
it to a problematic LibreSSL is a good idea. Sure, this means that |
| 47 |
some people will be forced to use LibreSSL because of libtls. |
| 48 |
|
| 49 |
But in practice, this means that any upstream starting to use libtls |
| 50 |
does a huge disservice to their users by preventing them from using |
| 51 |
their preferred SSL provider. So in the end, you either don't use |
| 52 |
libtls or your users are in pain. |
| 53 |
|
| 54 |
Actually, I see that someone has apparently forked libtls into libretls |
| 55 |
(the irony!) that works against OpenSSL [1]. |
| 56 |
|
| 57 |
[1] https://git.causal.agency/libretls/about/ |
| 58 |
|
| 59 |
|
| 60 |
-- |
| 61 |
Best regards, |
| 62 |
Michał Górny |
Archives