Re: Debian readline/libedit breakage

On 02/14/2011 08:27 AM, Greg Smith wrote:
> Markus Wanner wrote:
>> Anybody realized that this Debian bug (and several others) got closed in
>> the mean time (Sunday)? According to the changelog [1], Martin Pitt
>> (which I'm CC'ing here, as he might not be aware of this thread, yet)
>> worked around this issue by pre-loading readline via LD_PRELOAD for
>> psql.
>>
>> Personally, I'm a bit suspicious about that solution (technically as
>> well as from a licensing perspective), but it's probably the simplest
>> way to let only psql link against readline.
>
> This originated in
> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=608442 , and from
> what I'm reading there it sounds like Martin is inserting this as a
> workaround but it hasn't passed through Debian Legal yet. I would
> expect them to reject this as unacceptable. Dynamic linking via
> LD_PRELOAD is still linking, even if it happens at runtime. I commend
> Martin for buying some time here by doing that, but this doesn't
> change the urgency to come up with an alternate solution much to me.
> As I see it, that change could be reverted at any time via pushback
> from legal.
>
> As far as working around this by releasing our own packages goes,
> that's useful, but I'd also characterize that as only a workaround
> rather than a real solution. OpenSSL is open-source, but it's not
> "free software" via that standards of the FSF, which I feel is a
> completely reasonable position given the license. When you depend on
> a software stack built from unambiguously free software, having
> components that aren't you've wedged in there and are dependent on is
> never a good idea. I won't consider this truly resolved until GnuTLS
> support for PostgreSQL is in core.

Given the links Marko just posted, maybe NSS would be a better bet.
Apparently they also have some sort of compatibility library too.

I agree that the LD_PRELOAD trick seems absurd, and unlikely to be
acceptable to FSF types.

cheers

andrew