How well does the compactor handle large DNS messages, e.g. AXFR

[banburybill]

And how should it handle them?

At present I think a 300Mb AXFR will generate a few 300Mb buffers internally. This could be interesting.

[banburybill]

Currently you can select to ignore AXFR and IXFR. As things stand, this will record the AXFR query, but not the response RR, because we only check and ignore RR type on second and subsequent queries.

It would be better to move the Q/RRTYPE ignore into the packet decode. That way we can skip over large data we don't want. It would mean having to retain the original *COUNT values, as decoded packets would have fewer entries than the original.