todo

来自「Intercom 是一个 Unix系统上灵活的语音传输软件。支持标准音频压缩比如」· 代码 · 共 55 行

$Id: TODO,v 1.38 2002/10/06 04:51:02 shane Exp $

Work needing:

* Port to Windows.  This is non-trivial as cygwin only has a partial OSS
/implementation and we need a good low-level sound interface.

* Add call conferencing support.

* As it is now, several UDP packets usually come in to make a fragment.  We
force them to be sequential or they are dropped.  Rework this code as to
only drop packets if the sequence is wrong and a packet of a later sequence
has actually been written.  If packet(s) of a later sequence number are
still in the I/O buffers then we should accept the packet and insert it into
the buffer at the appropriate location. (not sure about importance,
implementation shouldn't be too hard if it's necessary at all).  Note: this
doesn't seem to ever happen in real-life.

* Add option to send each packet more than once.  Useful in cases where
there is high packet-loss.

* Parse command-aliases to expand $1, $2, ... into options given to that
alias.

* Drop silent mode and create noplay and nocapture mode respectively.  Also,
allow noplay and nocapture to be switched on during a call.  Note that doesn't
mean we close the dsp.

* Enable the use of or implement a directory service of some sort.  I'd
prefer to not re-invent the wheel here.

* Try to get piping or the like working.  "set |less" for example as it's
getting kinda long.  A pset command which runs set's output through less
would be fine as well I suppose.

* Implement proper handling of SIGSP signal.  At present, Intercom will most
likely crash if put into the background during a call.

* Have SIGINT terminate a call connection attempt if we are stuck in a
gethostname, connect, etc.  Actually, if a call is in progress, we could
just terminate the call, sigint twice would terminate Intercom.

* Add auto-gain-control using a compression algorithm of some sort.

* Could we convert silence_thresh to a decibel value so as to offer a
uniform interface?

* Implement some form of echo cancellation.  The basic idea is when the
remote party is sending audio above a certain threshold, we will stop
sending data of our own.  We could do it in reverse, when we are sending
data above say silence_thresh, we could cut reception if it was below
silence_thresh.  Quite clean actually but it would have the problem that the
data is being sent and thus doesn't same bandwidth.