Ashok, Please explain your comments below. They seem to be relative to making the change to fixed numbers.
This solves multiple problems 1) Minimize the changes in Base layer. ***The minimal change is no change, and the current approach (discovery based) works always. 2) Application does not require to store and retrieve the previously configured ports. ***The client application continues to use the previous port until it doesn?t work or client restarts. ***Then the client uses discovery to find the new port. Application doesn?t store anything. 3) If it fails to bind, anyway it has to retry other ports. ***Current approach never fails to bind. No need to ask for IANA numbers. Note that the previous approach used 10 assigned numbers, and I hope we won?t ask IANA for ten. John Light Intel OTC OIC development From: ASHOKBABU CHANNA [mailto:[email protected]] Sent: Tuesday, January 26, 2016 9:15 PM To: Light, John J; Uze Choi; 'Jon A. Cruz'; jaekeun lee; Macieira, Thiago; iotivity-dev at lists.iotivity.org; Markus Jung Subject: Re: Re: [dev] Need API to set static "sid" and "port number". Hi, To minimize the changes and also to increase the user experience, can we reserve a port with IANA for IoTivity? https://www.iana.org/form/ports-services. (Not sure who can apply for this) This solves multiple problems 1) Minimize the changes in Base layer. 2) Application does not require to store and retrieve the previously configured ports. 3) If it fails to bind, anyway it has to retry other ports. Previously we fixed 5298/6298 ports temporarily but as they are reserved, we moved to system assigned ports. As IoTivity supports multiple transports (BLE/BT), new API for only IP network will be strange for some developers. Regards, Ashok ------- Original Message ------- Sender : Light, John J<john.j.light at intel.com<mailto:john.j.light at intel.com>> Date : Jan 26, 2016 23:10 (GMT+05:30) Title : Re: [dev] Need API to set static "sid" and "port number". Uze Choi and Jon, I want to get on the table that implementing your proposal may be difficult and must be done carefully. The structure of IoTivity doesn?t make the change easy, and there are some possible pitfalls. First is that there are two ways to start IoTivity: OCInit() and OCInit1(). OCInit() is the original init, and it has the port argument you want. When I added IPv6, I needed additional arguments, so I added OCInit1(), with the needed arguments to allow configuring IPv6. Since this was an addition to the API, I asked for help from the maintainers, but received none, so it is my invention as it stands. Since I could see that some arguments of OCInit() were not used, I didn?t include them in OCInit1(). Among the removed arguments is the port number. If we upgrade OCInit() to use the port number, we need to also provide a way to do OCInit1() functionality with a port number so you can do the same thing with IPv6. Perhaps OCInit2()? The second issue is the startup path has some difficulties that will need to be worked out. New error codes will need to be added to address this issue, and the IP adapter will need to recognize the condition. The larger problem is that the semantics of starting up are not compatible now. Specifically, right now if any adapter starts up, errors in other adapters (including the one you want) are ignored. This existing behavior will have to be reconsidered. Finally, even if there is only one adapter (IP), OCSelectNetwork in ocstack.c doesn?t propagate error codes, returning only CA_STATUS_FAILED and CA_STATUS_OK. I still think this is a bad idea, but I wanted to provide information for a change if it is going to be done. John Light Intel OTC OIC development From: ???(Uze Choi) [mailto:[email protected]] Sent: Monday, January 25, 2016 9:49 PM To: Light, John J; 'Jon A. Cruz'; jjack.lee at samsung.com<mailto:jjack.lee at samsung.com>; Macieira, Thiago; iotivity-dev at lists.iotivity.org<mailto:iotivity-dev at lists.iotivity.org>; markus.jung at samsung.com<mailto:markus.jung at samsung.com> Subject: RE: [dev] Need API to set static "sid" and "port number". Hi Jon/John, I agree ?We know that the ?try? can be unsuccessful, so a discovery fallback is inevitable.? So that we need implement rediscovery UI for fallback case. However, if try is successful, we can reduce the step to re-discover which will bring the big benefit to user from UI perspective. How about including this issue into the Jira ticket for in this coming 1.1.0 release? Markus, could you help it? BR, Uze Choi From: iotivity-dev-bounces at lists.iotivity.org<mailto:iotivity-dev-bounces at lists.iotivity.org> [mailto:[email protected]] On Behalf Of Light, John J Sent: Tuesday, January 26, 2016 6:38 AM To: Jon A. Cruz; jjack.lee at samsung.com<mailto:jjack.lee at samsung.com>; Macieira, Thiago; iotivity-dev at lists.iotivity.org<mailto:iotivity-dev at lists.iotivity.org> Subject: Re: [dev] Need API to set static "sid" and "port number". Jon, I don?t see any attempt to propose static port binding, so the request doesn?t violate any OIC intent. In any case, since a node might support multiple servers, static port binding wouldn?t work. I understand the request is to make a server ?try? to use a port that it used in a previous run. We know that the ?try? can be unsuccessful, so a discovery fallback is inevitable. John Light From: Jon A. Cruz [mailto:[email protected]] Sent: Monday, January 25, 2016 1:30 PM To: Light, John J; jjack.lee at samsung.com<mailto:jjack.lee at samsung.com>; Macieira, Thiago; iotivity-dev at lists.iotivity.org<mailto:iotivity-dev at lists.iotivity.org> Subject: Re: [dev] Need API to set static "sid" and "port number". One aspect here sounds like code is depending on IoTivity devices to operate like some other common servers where HTTP is on port 80, telnet is on port 23, ssh on port 22, SMTP is on port 25, etc. Compliance with company IT policies, firewall requirements, vpn rules, etc. might come into play there. If this is so, then a question would be to ask if such port-bound services were counter to the OIC intent, within the OIC intent or un-addressed by it. On the other hand, if it is just code counting on finding a device by using address+port as a key, then I would share concerns on such software being fragile. On 01/22/2016 08:20 AM, Light, John J wrote: Jack Lee (?), In OCDoResource, you can supply a port number in the OCDevAddr argument. Network layer will use it, just as in original IoTivity. So the same functionality exists. I still have reservations about the design of counting on port numbers persisting. John From: iotivity-dev-bounces at lists.iotivity.org<mailto:iotivity-dev-bounces at lists.iotivity.org> [mailto:[email protected]] On Behalf Of ??? Sent: Thursday, January 21, 2016 9:47 PM To: Macieira, Thiago; iotivity-dev at lists.iotivity.org<mailto:iotivity-dev at lists.iotivity.org> Subject: Re: [dev] Need API to set static "sid" and "port number". - OCDoResource() must set port number. How can forget it? If you right, then iotivity should be redesign. Because, OCDoResource() has port number dependency. If we support just one optional api, we can reduce re-discovery problem. I don't understand why did yoy object to add "OPTIONAL" api although it can provide very useful user exprience. And, old version iotivity could set port number... there is no issue about development ------- Original Message ------- Sender : Thiago Macieira<thiago.macieira at intel.com<mailto:thiago.macieira at intel.com>> Date : 2016-01-22 14:19 (GMT+09:00) Title : Re: [dev] Need API to set static "sid" and "port number". On Friday 22 January 2016 02:25:47 jaekeun lee wrote: > I didn't say that change the mechanism. > I said, support the additional API for set to port number and sid or > di(variable name is sid) is very useful and helpful. > The experience depends on how the application is written > How can fix this issue? > - Presence Check -> Re-Discovery -> Check saved DI & discovered DI -> change > target's IP/PORT or > - re configure every run > These are very inconvenience . > > But, if server can set port, then re-configure only when server not work(may > be ip changed or port occupied). > Just additional API. Forget the port number. If your other side is trying to depend on the port number, please redesign it. You cannot count on the port number being free, so your other side must be able to deal with the port number changing. For that matter, the IP might change after a reboot. Or even without a reboot: IPv6 temporary addresses change about once a day. Since you can't count on the IP being fixed, port numbers don't identify anything. -- Thiago Macieira - thiago.macieira (AT) intel.com Software Architect - Intel Open Source Technology Center [cid:image001.gif at 01D158E7.A47C1520] [http://ext.samsung.net/mailcheck/SeenTimeChecker?do=28bccb0e7a1e0d063f4373047812202bf86cbca82d54dcb7a1889aa8efbbee47738ed17e7639ce1f641b1a8c451b073656239170f5eb4b5c326bbdfb2ea96a2fcf878f9a26ce15a0] -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.iotivity.org/pipermail/iotivity-dev/attachments/20160127/958cf948/attachment.html> -------------- next part -------------- A non-text attachment was scrubbed... Name: image001.gif Type: image/gif Size: 13168 bytes Desc: image001.gif URL: <http://lists.iotivity.org/pipermail/iotivity-dev/attachments/20160127/958cf948/attachment.gif>
