#grpc

[13grpc] 15jboeuf closed pull request #933: A couple of fixes + a test for client side authentication of hostnames (06master...06bugscrub2) 02http://git.io/x61j

[13grpc] 15tbetbetbe pushed 2 new commits to 06master: 02http://git.io/xXsc

13grpc/06master 141532923 15Jan Tattermusch: client side support for TLS

13grpc/06master 1493adddc 15Tim Emiola: Merge pull request #929 from jtattermusch/csharp_tls...

Remosi: gotcha, thanks

[13grpc-java] 15jcanizales opened pull request #170: Fixes javadoc errors (06master...06fixes-javadoc-installation-errors) 02http://git.io/xXr5

[13grpc] 15donnadionne opened pull request #940: Just a few more changes to include csharp into build and test (06master...06test-mod) 02http://git.io/xXDC

Anyone there ?

there's a few people, sure

what's up ?

great... I was just wondering if grpc can be used in conjunction with consul.io

somebody answered that a few hours ago actually :)

oh really. I waited for sometime and logged out ..

ah :D

so basically, yes, it is definitely possible to use grpc in conjunction with a service like consul.io, or any other

however, we haven't written anything for it yet

but there's nothing to prevent that from working

thats great . i should try it out then..

i was looking for some microservices frameworks for GO

if you manage to write some code to make that happen, well, contributions are welcome as github pull requests! :)

thanks..

[13grpc] 15jtattermusch pushed 2 new commits to 06master: 02http://git.io/xXS2

13grpc/06master 1438bf129 15Donna Dionne: Just a few more changes to include csharp into build and test

13grpc/06master 14513b9e3 15Jan Tattermusch: Merge pull request #940 from donnadionne/test-mod...

hello! since GRPC is embedded in HTTP/2, and it seems desirable to multiplex HTTP and RPC requests over one connection, are there plans to expose a simple HTTP interface that can share the channel?

[13grpc] 15ctiller pushed 2 new commits to 06master: 02http://git.io/x1Wc

13grpc/06master 14d30b005 15Nicolas "Pixel" Noble: Adding coverage report.

13grpc/06master 141acbf43 15Craig Tiller: Merge pull request #925 from nicolasnoble/coveralls...

@ashoemaker it's certainly possible but I suspect it's only relevant for contrained environments such as mobile

and possibly embedded devices

got a use-case in mind?

@louiscryan absolutely, the intended use case is mobile

sync & mutations over the RPC channel, content upload/download over HTTP/2 (avoid boxing bytes in protobufs, enable upload/download progress)

also resumable uploads

so you don't have to use protobufs with grpc, it would happily take raw chunked byte-sequences as the messages for resumable upload/download

youre going to want chunking for that

boxing that in protobuf will cost you about 4 bytes per chunk IIRC

I'm less concerned w/ the framing overhead, my concern w/ boxing reduces to: a) avoiding an extra copy during deserialization (maybe proto3 is more efficient?); b) providing upload/download progress.

however, if we can build on streams of chunked byte-sequences, then we can define a simple upload/download protocol that gives us progress information and resumability

for (b) C/C++ proto handles that pretty well, Java could use a little work IIRC

(b) duplex streaming is nice for progress & resumption as your service responds with chunk write ACKs

how were you handling progress indication over HTTP/1.1 ?

our current upload/download APIs indicate progress as bytes{Sent,Received} / bytesTotal

resumption is used only if the channel is disconnected, so I expect that we would need to restart the request after reconnecting

so this is full restart with no partial resumption?

we want to resume from the last ACKed offset – I see, you're pointing out that the write ACKs tell us exactly what that offset it

right

the partial resumption is optimistic of course, since the FE could have been restarted or we could be talking to a different FE if affinity is not preserved

with HTTP/1.1 you have to send individual requests for each chunk to get ACKs

that too

so this is what we do for similar use-cases at google

in HTTP/1.1 we'd need the extra roundtrip to determine where to resume, so the incremental ACKs are definitely superior

ah, makes sense

so all that said sharing the HTTP2 connection with a library that provides an HTTP/1.1 API should be doable

which language do you need?

are you thinking of mapping the GCS resumable upload API to GRPC?

C++ is ideal

our core library is shared between iOS and Android

and since our needs are so straightforward, we can layer upload/download over GRPC as-is if that works out nicely

that would be easiest at the moment

makes sense, and as you pointed out the chunk write ACKs have a clear advantage over an HTTP/1.1 interface

but we've definitely had ocnversations along these lines internally

e.g.

more specifically

in proto

rpc (stream WriteRequest) returns (stream WriteResponse)

where the WriteRequest contains the byte sequence and it's offset into the underlying source

and WriteResponse indicates the last durably written offset

that's great

using proto would require at least one extra copy, right?

not that it's a bottleneck of course

thanks @louiscryan, this was quite helpful. good night!

glad I could help,good night

[13grpc-java] 15ejona86 closed pull request #162: Add some details for building on Windows. (06master...06windows-worky) 02http://git.io/x2VB

[13grpc] 15yang-g opened pull request #947: Remove include google/protobuf/message.h (06master...06imacros) 02http://git.io/xSmy

[13grpc] 15jtattermusch opened pull request #949: fix gce script for cloud to cloud C# client (06master...06csharp_fix_gce_scripts) 02http://git.io/xSpU

[13grpc] 15donnadionne pushed 2 new commits to 06master: 02http://git.io/xSjL

13grpc/06master 14cfc7864 15Jan Tattermusch: fix gce script for cloud to cloud C# client

13grpc/06master 140431246 15donnadionne: Merge pull request #949 from jtattermusch/csharp_fix_gce_scripts...

[13grpc] 15nicolasnoble pushed 2 new commits to 06master: 02http://git.io/x9JT

13grpc/06master 14ee7ab3d 15Yang Gao: Remove include google/protobuf/message.h as it should be covered in config.h

13grpc/06master 14cac60b5 15Nicolas Noble: Merge pull request #947 from yang-g/imacros...