On Nov 7, 2007, at 11:29 PM, David Halik wrote:

> Jeff Johnson wrote:
>> Looks like rpm-4.4.9 was busted wrto network downloads, but
>> perhaps I deliberately changed. The issue of rpm accessing
>> a networks is somehow controversial even 7+ years later.
>>
>> Adding an explicit "r.ufdio" is as good as the other fix,
>> changing the default to make rpmio always network aware.
>> I still haven't identified why ftp access works on rpm-4.5,
>> but not in rpm-4.4.9, but I doubt its a solaris peculier issue.
>
> I'm quite happy with the quick fix at this point that's for sure.  
> We use the feature quite often in conjunction with network  
> installs, so broken network support was a big problem. I'm glad to  
> see that at least we can patch around it even if default support  
> isn't decided. I'll look at the other fix you suggested and maybe  
> just set it globally.... don't want your whiplash to get any worse ;)
>

;-) Actually the whiplash is mostly just a crick in the neck these days,
mostly that the current implementation http implementation does
not support redirects.

I'll likely replace the "r.ufdio" with a macro for each Fopen so that
users/distros can control whether, say, queries should be network
aware, but installs should not be permitted network access.

That implementation of policy is easier than rehashing the issue Yet  
Again.

>>
>> I'd suggest spending some time upgrading to rpm-4.5 as well. There  
>> should
>> be very very few surprises, and there were some significant  
>> cleanups wrto
>> removing void ** casts and more.
>>
>
> I was under the impression that 4.5 was still under development  
> since there hadn't been a full release, but if ftp is functional  
> out of the box I'll definitely check it out and throw it on our  
> test machines while waiting for a stable release. Are you planning  
> to release a 4.5 stable at this point or are you shooting right for  
> 5.0? Just something I've been wondering... I'm working towards a  
> complete conversion of our network for whenever 4.5 or 5 hits.
>

rpm-4.5 (and rpm-5.0) are both headed for release over the next 2  
months.

The primary goal for rpm-4.5 was to change the soname in rpm  
libraries, all of
which have an rpm version in order to establish a defacto API/ABI for  
the
existing rpm-4.4.9 release.

Another goal was to change paths to permit multiple versions of
rpm to be installed side-by-side if necessary.

And finally the intent for rpm-4.5 was to permit distros using  
rpm-4.5 to
establish what "everywhere compatible" really means, reverting any
changes and/or modifications since rpm-4.4.2 that were deemed
risky, unnecessary, misguided, or whatever.

The only development that has not been request driven since July
was speeding up queries by 10x 2 weeks ago (which also helped  
stabilize HEAD
at the time). But even that change will be reverted if desired.

So the rpm-4.5 branch code is largely rpm-4.4.9 cleaned up.

Ultimately, on Solaris, you likely want rpm-5.0, which has a  
considerably
more portable and maintainable build configuration than rpm has ever had
before.

>> Still no idea what the EPASV vs PASV problem root cause is.
>
> Is this actually a problem on your end? We are fairly certain that  
> our L4 is choking on the EPSV NAT translation vs. the more common  
> PASV, so this might not actually be an rpm issue, unless you saw  
> something otherwise. It looks like Tim was experiencing the same  
> ufdio/fdio issue and hadn't gotten to any FTP transfers (hence his  
> mention of ftpdebug not working). Jeff, you might have just tipped  
> us off on a networking issue specific to our case, which I thank  
> you for ;) hah.
>

Not a problem on my end, but Tim has said that a Solaris rpm client  
could not
access a RHEL vsftpd server.

But there are 2 problems, EPSV and non-functional ftp, that might  
explain
all the known issues.

I will likely have a reference build of rpm-4.5 on wraptastic.org  
this weekend
in order to start releasing rpm-4.5.

73 de Jeff