hylafax hickups - what's wrong here?

Question

We're running a hylafax server (6.0.6-7+deb9u1) installed on a raspberry pi (3), running raspbian stretch. The resulting fax-connections to our remote sides are unstable (about 3 of 10 faxes could not be sent at all / result in partial pages at the receiver side) but I cannot figure out why - cannot recognize any pattern.

In hope of getting a hint on where to search for a fix, I want to post a log of one of our problems ("no response to PPS repeat") here:

Hylafax-Log as GIST on github

I would be so happy if anyone could point me to the actual problem(s) - thank you so much in advance!

update:

In the meantime, we were able to setup another raspi behind another Fax-line and could send the same document with success - the log looks pretty much the same until lines 121 (error-log) and 153 (success-log) respectively:

Log of the successful run from another raspi

Answer 1

Posted the same question at the mailing list of hylafax, and got this response from Lee Howard:

---

The difference is the call route and the use of T.38 on the problematic call.

So, the problematic call experiences what HylaFAX sees as "poor line audio quality", and eventually one or both endpoints give up retrying.

The resolution would be to change phone service providers, disable T.38, and otherwise make attempts to remove audio quality problems.

---

Furthermore, he commented the reversed NSF remote station ID with:

---

It's because the T.38 gateway being used is trying to obfuscate the NSF in order to prevent the sender from attempting to negotiate non-standard features with it.

Notice the difference:

Jun 29 09:57:51.03: [ 902]: REMOTE NSF "00 00 00 48 79 6C 61 46 41 58 20 28 74 6D 29 20 56 65 72 73 69 6F 6E 20 35 2E 36 2E 31" Jun 29 09:57:51.03: [ 902]: NSF remote fax equipment: unknown - indeterminate Jun 29 09:57:51.03: [ 902]: NSF remote station ID: "1.6.5 noisreV )mt( XAFalyH"

... versus...

Jun 29 10:34:10.55: [24880]: REMOTE NSF "AD 00 55 48 79 6C 61 46 41 58 20 28 74 6D 29 20 56 65 72 73 69 6F 6E 20 35 2E 36 2E 31" Jun 29 10:34:10.55: [24880]: NSF remote fax equipment: HylaFAX Jun 29 10:34:10.55: [24880]: NSF remote station ID: "HylaFAX (tm) Version 5.6.1"

In the first example the T.38 gateway being used changed the first three bytes of the NSF signal to "00 00 00" instead of their correct "AD 00 55".

This is common but unfortunate behavior from T.38.