use full range for sequence numbers #374
Conversation
additionally, remove unused type ethif from Arpv4 module
hannesm
force-pushed
the
sequence-full-range
branch
from
408e151
to
feca22b
Compare
…res ocaml 4.03, which provides result in Pervasives)
hannesm
force-pushed
the
sequence-full-range
branch
from
feca22b
to
0397b6c
Compare
hannesm
force-pushed
the
sequence-full-range
branch
from
0397b6c
to
a294769
Compare
| let localport = | ||
| 10000 + (Randomconv.int (fun x -> Random.generate x) ~bound:10000) | ||
| 1024 + (Randomconv.int ~bound:(0xFFFF - 1024) Random.generate) |
There was a problem hiding this comment.
This patch looks like a definite improvement to me!
Out of curiosity, do we have to choose local ports > 1024? Is this a Unix convention or a TCP convention? (No big deal, I'm just interested)
There was a problem hiding this comment.
oh, of course any port number is fine, according to the RFCs[tm]. in practise, there are servers which refuse to service you if your source port range is < 1024 (I experienced this with some authoritative nameservers) - thus I'd (for now at least) avoid to use something < 1024 -- but then our UDP implementation takes any source port -- maybe we should unify these two expressions!?
|
related to this is hannesm@562e6e4 together with hannesm/mirage@a24e88a and hannesm/charrua@56be972 and hannesm/mirage-qubes@4286ee7 -- which adds RANDOM (and MCLOCK, see #373 -- I want to avoid to have to do the same set of releases again next week just because I didn't integrate MCLOCK yet) to the Static_ipv4 interface. if someone could tell me that's the right direction or not, I'm happy to incorporate these 4 commits tomorrow in the (anyway) pending releases (towards 3.2.0) /cc @mirage/core the issue at hand is: the IPv4 stack (Static_ipv4, via Ipv4_common) still uses OCaml's Random number generator, while we have this nice Mirage RANDOM device, which can be used either with stdlib-random or fortuna/nocrypto. |
|
included in #371, together with the mentioned IPv4 functor adjustment. |
(on top of #371, relevant commit is only the last one)