Corporate firewall braindamage
The apparent commonality of corporate firewall braindamage, and the resulting "need" of people to pull over dumb (http) transport, is an ongoing problem on kernel.org.
I have thought some about what can be done to improve the situation, and I have come up with the following list of possibilities, pretty much listed in order from easiest and least generic to hardest and most generic.
It would be very interesting if people who have familiarity with this particular class of braindamaged firewalls could comment on how many users would be helped by which ones of these solutions.
1. git protocol via CONNECT http proxy
Connect to http proxy, and use a CONNECT method to establish a link to the git server, using the normal git protocol.
Minor change to TCP connection setup, but no other changes needed. No changes on the server side.
2. git protocol over SSL via CONNECT http proxy
Same as #1, but encapsulate the data stream in an SSL connection. If the git server is run on port 443, then the fact that the data on the SSL connection isn't actually HTTP should be invisible to the proxy, and thus this *should* work anywhere which allows https://
traffic.
Requires the git server to speak SSL.
3. git protocol encapsulated in HTTP POST transaction
git protocol is already fundamentally a RPC protocol, where the client sends a query and the server responds. Furthermore, it tries to minimize the number of round trips (RPC calls), which is of course desirable.
Each such RPC transaction could be formulated as an HTTP POST transaction.
This requires modifications to both the client and the server; furthermore, the server can no longer rely on the invariant "one TCP connection == one session"; a proxy might break a single session into arbitrarily many TCP connections.
Thoughts?
-hpa
Quoted 7 lines · H. Peter Anvin
1. git protocol via CONNECT http proxy
Connect to http proxy, and use a CONNECT method to establish a link to the git server, using the normal git protocol.
Minor change to TCP connection setup, but no other changes needed. No changes on the server side.
Many firewalls will detect that CONNECT will not going to 443 and block you, and even if you run git:// daemon on 443, they will detect that you are not talking SSL initial exchange and shut you off.
Quoted 9 lines
2. git protocol over SSL via CONNECT http proxy
Same as #1, but encapsulate the data stream in an SSL connection. If the git server is run on port 443, then the fact that the data on the SSL connection isn't actually HTTP should be invisible to the proxy, and thus this *should* work anywhere which allows https://
traffic.Requires the git server to speak SSL.
Yes, perhaps putting it behind an independent ssl relay would give you a solution without any code change.
Quoted 7 lines
3. git protocol encapsulated in HTTP POST transaction
git protocol is already fundamentally a RPC protocol, where the client sends a query and the server responds. Furthermore, it tries to minimize the number of round trips (RPC calls), which is of course desirable.
Quoted 7 lines
Each such RPC transaction could be formulated as an HTTP POST transaction.
This requires modifications to both the client and the server; furthermore, the server can no longer rely on the invariant "one TCP connection == one session"; a proxy might break a single session into arbitrarily many TCP connections.
It would probably be a one-CS/EE-student-half-a-summer sized project to create such a server-side support with a specialized client.
Quoted 11 lines · Junio C Hamano
Quoted 10 lines · H. Peter Anvin
3. git protocol encapsulated in HTTP POST transaction
git protocol is already fundamentally a RPC protocol, where the client sends a query and the server responds. Furthermore, it tries to minimize the number of round trips (RPC calls), which is of course desirable.
Each such RPC transaction could be formulated as an HTTP POST transaction.
Quoted 7 lines
Quoted 4 lines
This requires modifications to both the client and the server; furthermore, the server can no longer rely on the invariant "one TCP connection == one session"; a proxy might break a single session into arbitrarily many TCP connections.
It would probably be a one-CS/EE-student-half-a-summer sized project to create such a server-side support with a specialized client.
Funny you say that. This was a GSoC 2008 project idea. We even received an application from a student for it.
The hard part is either making the server side stateful, so it can remember what the last RCP call had said it wants/haves, or doing a stateless protocol where the client uses an exponential expansion (or some such behavior) of its have list until the server replies with the pack data.
Signature · 1 line
Shawn.
Quoted 10 lines · Shawn O. Pearce
Funny you say that. This was a GSoC 2008 project idea. We even received an application from a student for it.
The hard part is either making the server side stateful, so it can remember what the last RCP call had said it wants/haves, or doing a stateless protocol where the client uses an exponential expansion (or some such behavior) of its have list until the server replies with the pack data.
One easy way of doing the former is to have a session reassociator in the flow; pretty much a multiplexer which receives the HTTP request, and passes it onto a work slave (which can be an ordinary process, in fact, can be the ordinary git daemon) based on a session and sequence ID.
-hpa
Quoted 5 lines · H. Peter Anvin
Quoted 4 lines · Shawn O. Pearce
Funny you say that. This was a GSoC 2008 project idea. We even received an application from a student for it.
Quoted 9 lines
Quoted 6 lines
The hard part is either making the server side stateful, so it can remember what the last RCP call had said it wants/haves, or doing a stateless protocol where the client uses an exponential expansion (or some such behavior) of its have list until the server replies with the pack data.
One easy way of doing the former is to have a session reassociator in the flow; pretty much a multiplexer which receives the HTTP request, and passes it onto a work slave (which can be an ordinary process, in fact, can be the ordinary git daemon) based on a session and sequence ID.
s/multiplexer/demultiplexer/
The best might be to turn the demultiplexer either into an Apache module or some scripting language which can run inside Apache (e.g. mod_perl) to avoid Apache spawning a CGI program which is only used to talk to the git daemon backend.
-hpa
Quoted 22 lines · Junio C Hamano
Quoted 7 lines · H. Peter Anvin
1. git protocol via CONNECT http proxy
Connect to http proxy, and use a CONNECT method to establish a link to the git server, using the normal git protocol.
Minor change to TCP connection setup, but no other changes needed. No changes on the server side.
Many firewalls will detect that CONNECT will not going to 443 and block you, and even if you run
git://daemon on 443, they will detect that you are not talking SSL initial exchange and shut you off.Quoted 8 lines
2. git protocol over SSL via CONNECT http proxy
Same as #1, but encapsulate the data stream in an SSL connection. If the git server is run on port 443, then the fact that the data on the SSL connection isn't actually HTTP should be invisible to the proxy, and thus this *should* work anywhere which allows https://
traffic.
Quoted 4 lines
Quoted 1 line
Requires the git server to speak SSL.
Yes, perhaps putting it behind an independent ssl relay would give you a solution without any code change.
in more pananoid locations they are putting client certs on desktops and giving those to the IDS systems so that they can decrypt the SSL traffic, so if it doesn't look like HTTP inside the SSL they will block it.
this isn't very common now, but the firewalls that are blocking #1 weren't very common a year or so ago either.
Quoted 10 lines
Quoted 10 lines
3. git protocol encapsulated in HTTP POST transaction
git protocol is already fundamentally a RPC protocol, where the client sends a query and the server responds. Furthermore, it tries to minimize the number of round trips (RPC calls), which is of course desirable.
Each such RPC transaction could be formulated as an HTTP POST transaction.
Quoted 7 lines
Quoted 4 lines
This requires modifications to both the client and the server; furthermore, the server can no longer rely on the invariant "one TCP connection == one session"; a proxy might break a single session into arbitrarily many TCP connections.
It would probably be a one-CS/EE-student-half-a-summer sized project to create such a server-side support with a specialized client.
this is probably the best long-term option.
David Lang