Previously we had a confusing situation, with:
* single-arg doc: published name <name> to <value>
* double-arg doc: published name <value> to <name>
* implementation: Published name <name> to <value>
Now we have the uniform:
Published to <name>: <value>
With the following goals:
1. It's clear that we're writing <value> to <name>'s IPNS slot in the
DHT.
2. We preserve the order of arguments from the command-line
invocation:
$ ipfs name publish <name> <value>
Published to <name>: <value>
This lets users resolve (recursively or not) DNS links without pulling
in the other protocols. That makes an easier, more isolated target
for alternative implemenations, since they don't need to understand
IPNS, proquint, etc. to handle these resolutions.
For explicitly enabling recursive behaviour (it was previously always
enabled). That allows folks who are interested in understanding
layered indirection to step through the chain one link at a time.
This allows direct access to the earlier protocol-specific Resolve
implementations. The guts of each protocol-specific resolver are in
the internal resolveOnce method, and we've added a new:
ResolveN(ctx, name, depth)
method to the public interface. There's also:
Resolve(ctx, name)
which wraps ResolveN using DefaultDepthLimit. The extra API endpoint
is intended to reduce the likelyhood of clients accidentally calling
the more dangerous ResolveN with a nonsensically high or infinite
depth. On IRC on 2015-05-17, Juan said:
15:34 <jbenet> If 90% of uses is the reduced API with no chance to
screw it up, that's a huge win.
15:34 <wking> Why would those 90% not just set depth=0 or depth=1,
depending on which they need?
15:34 <jbenet> Because people will start writing `r.Resolve(ctx, name,
d)` where d is a variable.
15:35 <wking> And then accidentally set that variable to some huge
number?
15:35 <jbenet> Grom experience, i've seen this happen _dozens_ of
times. people screw trivial things up.
15:35 <wking> Why won't those same people be using ResolveN?
15:36 <jbenet> Because almost every example they see will tell them to
use Resolve(), and they will mostly stay away from ResolveN.
The per-prodocol versions also resolve recursively within their
protocol. For example:
DNSResolver.Resolve(ctx, "ipfs.io", 0)
will recursively resolve DNS links until the referenced value is no
longer a DNS link.
I also renamed the multi-protocol ipfs NameSystem (defined in
namesys/namesys.go) to 'mpns' (for Multi-Protocol Name System),
because I wasn't clear on whether IPNS applied to the whole system or
just to to the DHT-based system. The new name is unambiguously
multi-protocol, which is good. It would be nice to have a distinct
name for the DHT-based link system.
Now that resolver output is always prefixed with a namespace and
unprefixed mpns resolver input is interpreted as /ipfs/,
core/corehttp/ipns_hostname.go can dispense with it's old manual
/ipfs/ injection.
Now that the Resolver interface handles recursion, we don't need the
resolveRecurse helper in core/pathresolver.go. The pathresolver
cleanup also called for an adjustment to FromSegments to more easily
get slash-prefixed paths.
Now that recursive resolution with the namesys/namesys.go composite
resolver always gets you to an /ipfs/... path, there's no need for the
/ipns/ special case in fuse/ipns/ipns_unix.go.
Now that DNS links can be things other than /ipfs/ or DHT-link
references (e.g. they could be /ipns/<domain-name> references) I've
also loosened the ParsePath logic to only attempt multihash validation
on IPFS paths. It checks to ensure that other paths have a
known-protocol prefix, but otherwise leaves them alone.
I also changed some key-stringification from .Pretty() to .String()
following the potential deprecation mentioned in util/key.go.
This should fix issue #1196 (Can't launch a command line
process from Qt).
The check was bad because it took stdin into account,
but it really shouldn't.
License: MIT
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
This should fix issue #1141 (ipfs cat "multihash too short"
error when using stdin) and perhaps others.
License: MIT
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
After losing jenkins, it's been difficult to test all commits
manually. This commit adds a Makefile target that makes travis do it.
Unfortunately, this is way too slow. It takes longer than the
allotted 10min.
After asking the travis people what to do, someone suggested making
sure that each commit is pushed to github independently. This makes
travis run CI on every single commit in the PR, and gives us nice
status indicators on each one (so we know which ones did not pass).
This approach means that we need to push a branch to the repo for
each commit in the PR-- otherwise travis may cancel its run if it
detects that the branch is no longer there. We could automate this
with a bot that essentially does:
for each PR:
git fetch the PR branch
push a branch per commit: <branch>-<commit>
for each closed PR:
delete all branches with pattern <branch>-<commit>
We had a very nasty problem: handshakes were serial so incoming
dials would wait for each other to finish handshaking. this was
particularly problematic when handshakes hung-- nodes would not
recover quickly. This led to gateways not bootstrapping peers
fast enough.
The approach taken here is to do what crypto/tls does:
defer the handshake until Read/Write[1]. There are a number of
reasons why this is _the right thing to do_:
- it delays handshaking until it is known to be necessary (doing io)
- it "accepts" before the handshake, getting the handshake out of the
critical path entirely.
- it defers to the user's parallelization of conn handling. users
must implement this in some way already so use that, instead of
picking constants surely to be wrong (how many handshakes to run
in parallel?)
[0] http://golang.org/src/crypto/tls/conn.go#L886
