>>Gnutella (when I tried it) seems to have about 10x as much as napster
>>in terms of currently online files (10 Tb when I checked over 2500
>>nodes). On Gnutella people are sharing not just mp3s but music videos
>>in the 20-80Mb size range, and even what looked like a movie of 350 Mb.
>>
>>Gnutella stands a much better chance of success because napster
>>servers being central can be closed down. Gnutella basically can't be
>>shutdown.
Ten Terabytes - Good Grief! Mighty large haystack to hide needles in.
Somebody has suggested that Metallica should set up a few thousand
high-bandwidth Napster connections, and start flooding Napster with
MP3s that have their song names but instead contain a recorded
"Hey, you, stop ripping off our music!" message...
The real problem with Gnutella and FreeNet is that they don't
provide a good mechanism for finding other users near you to
conserve bandwidth - leading to the problem of university networks
getting swamped by students downloading on limited internet feeds
instead of staying on the campus LAN, getting them from other students.
Napster was able to reduce this problem significantly by modifying their
indexing mechanism, so students would get copies from inside,
but systems that hide where servers are to prevent censorship
and don't have central databases to track things are very bad at this.
It's sometimes possible to build useful indexing on top of them,
but it's much harder if it's not planned for from the beginning.
Pointcast was a news-broadcasting screensaver that got into this
kind of trouble when everybody started using it, because it
didn't take advantage of web caching. After they got blocked from
a large number of corporate networks, they semi-fixed the problem by
building a cache/relay server companies could buy and deploy inside,
but it's tougher to sell that for a copyright-evasion server.
Anybody have creative suggestions for fixing them?
- It's easy to hide Gnutella-like things from sysadmins -
either build them to look like existing protocols,
or build them to use random port numbers so they're hard to block,
but then you lead to connection-swampage and congestion collapse.
It's much better to find a way to make the applications behave non-rudely
than to hide them where they're hard to block except by mass witchhunts.
- But if you're trying to hide where you are, it's hard to ask who's
nearby....
- It's possible to have indexes without central servers - you distribute
indexes to stuff using the same mechanism you distribute files,
and try to find a way to locate and update indexes.
For instance, maybe distribute indexes to your files,
or maybe distribute an updatable index to Non-Metallicoids's
along with the songs, and make the keyword "index" searchable.
- Round-trip times give you some hints for localized fetching -
but also give out hints on where you are.
- It's much harder to hide from the KGB than the Copyright Police,
because the latter may be fanatics but have limited resources.
- You need to design the system to be resistant against
people trying to break the whole thing. For instance, some flamer
killed off a lot of remailers by remailering slander about himself
to Usenet and then complaining the the ISPs where the remailers were.
Unless Napster wins tolerably well, a few dozen big bands suing them
could
Thanks!
Bill
Bill Stewart, [EMAIL PROTECTED]
PGP Fingerprint D454 E202 CBC8 40BF 3C85 B884 0ABE 4639
