Since the streaming servers' addresses will never show to users unless they use a network monitor such as Firebug, the address does not matter at all, as long as it's static.
Very far, or not far, depending on tools.
If we can use e.g. mencoder (which emits somewhat broken FLV sometimes), this kind of script can be used completely automatically to convert an AVI into FLV1 (H263 video):
mencoder \
input.avi \
-mc 0 \
-o output.flv \
-of lavf -lavfopts format=flv \
-oac mp3lame -af resample=22050 -lameopts preset=30:aq=0 \
-ovc lavc \
-lavcopts vcodec=flv:vqscale=9:mbd=2:trell:cbp:v4mv:cmp=6:precmp=3:dia=-3:predia=-4:vb_strategy=1:preme=2:vpass=1
for s in 2 3 4 5; do
mencoder \
input.avi \
-mc 0 \
-o output.flv \
-of lavf -lavfopts format=flv \
-oac mp3lame -af resample=22050 -lameopts preset=30:aq=0 \
-ovc lavc \
-lavcopts vcodec=flv:vqscale=11:mbd=2:trell:cbp:v4mv:cmp=6:precmp=3:dia=-3:predia=-4:vb_strategy=3:preme=2:vpass=3
done
# This puts an index into the flv, for seekability:
wine ./flvmdi.exe output.flv /k
It preserves the movie's length and whole content including any subtitles, since it is made by re-encoding from the published AVI.
If sufficient software can be installed on the/a TASVideos server to do this, it could do it automatically for any published AVI once it has been downloaded.
Then uploading it to the servers ― I was thinking maybe the server administrators could rather download the FLVs themself, so they could decide which movies they particularly want to stream…
30 FPS (using whatever technique somebody found for smoothing the video frames) would probably look best. Less bandwidth eaten, for one.
I would consider doing it, but would there be any rewards for this? Bandwidth and server space isn't free.