[tor-project] Onion sites vs onion services vs hidden services

Roger Dingledine arma at mit.edu
Wed May 4 09:02:14 UTC 2016

On Wed, May 04, 2016 at 07:34:38AM +0000, Colin Childs wrote:
> Just to clarify, the feedback I received was that "Onion Service" is
> generally the preferred term; however "Onion site" could be used as a
> sub-class of "Onion Service" to refer exclusively to websites. Due to
> this, the user manual[1] has been changed to use the term "Onion Service".

Right. Onion site is a fine name for an onion service that is served
by a webserver (akin to the word 'website').

The phrase onion site, like the word website, reinforces the "consumer"
model of the Internet. Onion services are inherently better at being
peer-to-peer (that is, bidirectional rather than unidirectional)
compared to ordinary Internet services, because they ignore NAT, and
because their addressing scheme is independent from static IP addresses.

As Yawning pointed out, Ricochet is a great example of an onion service,
but there's no webserver serving pages, so my Ricochet client is not an
onion site.

At least, that's the terminology we seem to have converged on. There
is still time to change it if people feel strongly enough.

> That said, we are still very inconsistent about how we use these terms
> in many other places.

Agreed. I had an idea to do the terminology switchover on the Tor website
just before the 32c3 onion services talk, and then announce it there.
I didn't get enough momentum to do it then, but I'm still a fan.


More information about the tor-project mailing list