With git, you can forget to push after a commit, but in practise, I have forked and pulled multiple branches without issue.
River's current development model fits well with Git - fork and pull. River developers wanted a stable trunk, with separate development branches that are later merged, the last merge was a difficult and error prone process. SVN does have better provenance, it's probably preferable to git when trunk is the main development branch which goes through a stabilisation period before branching for each release.
It'll be much easier to fork and pull my new code once we change to git.With the current policy, I first need to branch trunk (an earlier branch from where I branched) to skunk, commit my current work to skunk, then merge that back with trunk.
In any case, not moving to git is not a show stopper, but it does make life harder with our current development model.
Cheers, Peter. On 12/11/2016 6:30 AM, Bryan Thompson wrote:
I do not thing so. The point is to minimize the complexity for the people doing the import by breaking things up first so it can go into separate git repos immediately. Bryan On Fri, Nov 11, 2016 at 10:38 AM, Patricia Shanahan<p...@acm.org> wrote:I am a little troubled by the need to do the rearranging in svn before copying to git. That seems to imply svn has better reorganization features. On 11/11/2016 3:22 AM, Peter wrote:We've got a little work to do. Cheers, Peter.[https://issues.apache.org/jira/browse/INFRA-12432?page=com. atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Lambertus updated INFRA-12432: ------------------------------------ Status: Waiting for user (was: Waiting for Infra) This will be complicated and time consuming as the svn repo doesn't fit the usual trunk/branches/tags format. You may want to do some consolidation or break this down into multiple git repos, for example, river-site.git, river-rt-tools.git, etc. prior to us doing an SVN->Git migration. If you'd rather have it all in one repo, let us know and we'll see what we can do.Apache River migration from SVN to Git - Git commit access---------------------------------------------------------- Key: INFRA-12432URL:https://issues.apache.org/jira/browse/INFRA-12432Project: Infrastructure Issue Type: New Git Repo Components: Git Reporter: Peter Firmstone-- This message was sent by Atlassian JIRA (v6.3.4#6332)On 11/11/2016 3:23 PM, Niclas Hedhman wrote:File a JIRA ticket on INFRA project should work. Specify which SVN subtree to migrate into each repository. On Fri, Nov 11, 2016 at 1:15 PM, Patricia Shanahan<p...@acm.org> wrote: What is the current process for getting a writable GIT repository withinASF? On 11/10/2016 8:25 PM, Peter wrote: Thinking about how to proceed with code and repo...* I want to bring code in from an existing git repo and keep commit history (I'm the only committer). Branched off a recent version of River trunk. * We want to change to using a git repo for River. * Start building maven style modules from the ground up, starting with JERI at low level. * Separate out the qa test suite (integration tests), which is an ant build that only depends on jars from river build. * Where should jtreg test suite ( unit and regression tests) go? Maybe with each relevant module? * junit tests with appropriate module... Thoughts / suggestions? Regards, Peter. Sent from my Samsung device. Include original message ---- Original message ---- From: Peter<j...@zeus.net.au> Sent: 06/11/2016 08:23:06 pm To: dev@river.apache.org Subject: Re: River revamp Yes same pattern, some details are different for security reasons, additional support is required for lower level protocols as object endpoints. Also, for example, devices were on a 6LowPAN network, different types of devices would subscribe to different multicast groups, to minimise their responses, since some devices may rely on battery power, we don't want them to announce their presence or respond unnecessarily as that wastes power. There are a number of new IoT protocols being developed, I think we'll need to provide some support for some to start with. AMQP is another interesting protocol. On the discovery side, in order to make a connection, the multicast response will need to contain the application layer name, transport layer name, contact address and port, eg: MQTT, TCP, IP address:port. Typically the nework layer will define the method of multicast. Cheers, Peter. On 6/11/2016 1:05 PM, Niclas Hedhman wrote: Ok, so this is more or less classic "surrogate" setup with JINI,right, with some additional SEC stuff, right? And that is a cool way to achieve interoperability with smaller devices without ability to run a JVM, especially the original dream where devices don't know about each other ahead of time (except by some interface) I also see value where "IoT Service" doesn't bother with "Service Registrar" in the "Jini sense" and the "IoT Device" only participate in "Discovery" and then get a secure/trusted channel, onto which a light-weight protocol, such as MQTT or CoAP, can be funneled in either direction. On Sat, Nov 5, 2016 at 2:26 PM, Peter<j...@zeus.net.au> wrote: Hmm lets try Ascii, hope line wrapping doesn't wreck it.|<----------------------| Multicast request Multicast | | response |---------------------->| Connection& auth | | to discovered address RPCSEC_GSS |<----------------------| auth |---------------------->| RPCSEC_GSS Auth | | success, request | | bytes containing | | service proxy | Register service | | proxy& attributes |------------------->| Registration | Mange reg lease |<-------------------| | | | | | Match |<-----------------| Lookup | | |----------------->| | | | | Authenticate Iot Service | | | | with bootstrap proxy | | | | Grant permission to download | Auth client |<---------------------------- ----------| and deserialize service proxy. | Return service proxy |----------------------------- --------->| | | | | Prepare service proxy |<---------------------------- ----------------------------------| execute RPC function call Function |---------------------------- ---------------------------------->| with constraints | | | | IoT Device IoT Service Service Registrar Client On 5/11/2016 3:48 PM, Peter wrote: Ok, will come back to the ascii art, I'll first attempt to attach apng. There's an existing Java RPC implementation LGPL, with a maven build tag => oncrpc4j-2.6.0 https://github.com/dCache/oncrpc4j The implementation above supports /RPCSEC_GSS/ for security, although there's a client side bug at present, but fixing it will be a lot less work than reimplimenting it. Basically RPC is the C equivalent of Java RMI. Cheers, Peter. On 5/11/2016 10:32 AM, Niclas Hedhman wrote: Sorry, I get the feeling that too much detail thinking is still inyour head. Hard for me to follow your thought process. A simple picture (ascii art would do) would go a long way... Niclas On Sat, Nov 5, 2016 at 8:04 AM, Peter<j...@zeus.net.au> wrote: Thinking about C and power constrained devices, how about the following: * Write an ONC RPC java compiler, to create java code(instead of C) client stubs. * Provide support for TLS and constraints. * Provide an IPv6 constrained device announcement (C) and discovery (Java) utilities. Create a standard so other languages can be supported by others. * Write a java utility and service that manages proxies, registers discovered constrained devices with a lookup service and manages it's lease. This utility can generate attributes (from Configuration) and provide a bootstrap proxy (service) to allow clients to authenticate and obtain the smart proxy used to communicate directly with the device. * Provide an interface for clients to notify the utility service when a device is down. Regards, Peter. Sent from my Samsung device. Include original message ---- Original message ---- From: Zsolt Kúti<la.ti...@gmail.com> Sent: 03/11/2016 05:37:45 pm To: dev@river.apache.org Subject: Re: River revamp A small footprint implementation of Jini's lookup service written in C, fully JCK compliant. http://www.psinapticcom/link_files/PsiNapticTelematics.pdf A few years ago being involved in developing a streelighting management system I tried to access them to no avail. On Thu, Nov 3, 2016 at 8:09 AM, Peter<j...@zeus.net.au> wrote: I've been conaidering that. It should be possible to implement service discovery in C, any serialized java bytes required for a proxycould be stored on the device. So these devices are services, but not clients. Will respond with more soon Regards, Peter. Sent from my Samsung device. Include original message ---- Original message ---- From: Niclas Hedhman<nic...@hedhman.org> Sent: 03/11/2016 12:39:33 pm To: dev@river.apache.org Subject: Re: River revamp "IoT" is a term that for this discussion is a bit too wide. The "thermostat" runs with a kB-sized microcontroller and is struggling to getsecurity features in at all, and the "home router" is typically(still) running from a 4-8MByte flash, which is impossible to even get a Java ME onto, so there is a lot of challenges when using "IoT" as a blanket term.So, I think a couple of concrete, do-able, use-casesneed to be highlighted as examples, maybe a kind of "blue print" paper on how to do it with River. I totally agree that the "mothership" model is outrageous from a consumer's perspective, a nasty vendor lock-in, that all vendors are pushing for andall consumers/users need to fight the best we can.A very active home automation project is called "OpenHAB", a flurry of activity, connecting just about everything from your thermostat to your dog's toys. I have not looked closely at it, don't even know if it is a Java project as such, but it is one of the most active projects in the field of Home IoT. Cheers Niclas On Wed, Nov 2, 2016 at 10:41 PM, Patricia Shanahan<p...@acm.org> wrote: I think for this to work it is necessary to find out where IoT people hang out, and discuss it there Do they already have a plan for a secureframework? As a potential IoT user I'm looking for two things: 1. Security. 2. Server independence. I don't want my thermostat to stop working if a server I don't control goes down or is taken out of service for any reason. I think River may be a good basis for those features. On 11/2/2016 7:22 AM, Bryan Thompson wrote: I look at this as open source for secure IoT. The need for security in IoThas been aptly demonstrated by recent DDOSattaches based on compromiseddevices. I do feel that interop is critical to success here.Do we have any lurkers from the IoT manufacturing space here? People orcompanies willing to invest time and resources for a secure IOT platform? Bryan On Wednesday, November 2, 2016, Peter<j...@zeus.net.au>wrote: Utilising most of the existing discovery code, we could use ipv6 multicast, for an exported remote object (service).Then create a new class called RemoteDiscovery to discover a service dynamically, based on a name So you export a service and it becomes dynamically discoverable. It's not going to step on any Jini discovery lookup stuff and it's goingto be easily deployed by new users.Then once users realise there's more on offer theycan take advantage astheir understanding develops.Cheers,Peter. Sent from my Samsung device. Include original message ---- Original message ---- From: Niclas Hedhman<nic...@hedhman.org<javascript:;>> Sent: 02/11/2016 05:31:26 pm To: dev@river.apache.org<javascript:;> Subject: Re: River revamp To put a bit more meat on Peter's condensed list... I put forward a proposal to sever the ties between River and Jini itself,and instead re-focus River to be a a secured networktransport, with optional discovery. Starting point is of course the JERImodule and Peter's work to secure this transport, but in the longer term look at alternativetransport formats and eventually bindings to otherlanguages, which I thinkwill be the major hurdle for long term acceptance (no one is Java- onlynowadays).Jini's services, Reggie and so on, carries a lot ofnegative connotation among people who were around back then, and exceptfor where it has beenadopted, I doubt that there will be any new uptake,so instead of makingJini (and its specs) the focal point of River, make itto "Examples of whatRiver can be used for". Another example of what can be done with River could eventually includeconnectors for popular platforms, such as Zookeeper,which could openavenues for new blood coming to RiverConcrete things; Apache Karaf is also a very smallcommunity, yet they havemanaged to put together a very exciting website, and I think River community could "borrow" a lot of that work, making itself more appealing, promoting the new focus. I don't think much coding is needed to get thisgoing, but packaging might be "fixed" to makeconsumption of the corefunctionality as easy as possible, preferably easier thanthat. Once that is up-and-running starting the "reachout" to other projects, individuals and press releases.I hope that this will inspire some to more action.Niclas On Wed, Nov 2, 2016 at 2:25 PM, Peter Firmstone< peter.firmst...@zeus.net.au<javascript:;> wrote: A discussion recently ignited on river privateabout revamping the project.For the benefit of the wider developer community can we restate the suggestions here, feel free to reword, correct, reject orsuggest It was along the lines of:* Website revamp* Remove Jini focus, with a historical section... * Focus on new security features. * Make getting started simple, with just the bare bones basics, Extensible remote invocation with secure serialization.* Services, Javaspaces etc, become examples of whatcan be done withRiver, not what River is.Regards,Peter.Sent from my Samsung device. --Niclas Hedhman, Software Developer http://zest.apache.org - New Energy for Java --Niclas Hedhman, Software Developerhttp://zest.apache.org - New Energy for Java
