Re-, The port sets should not overlap; this is a trivial assumption of this work.
Cheers, Med ________________________________ De : Behcet Sarikaya [mailto:sarikaya2...@gmail.com] Envoyé : mercredi 21 septembre 2011 08:42 À : BOUCADAIR Mohamed OLNC/NAD/TIP Cc : Softwires-wg Objet : Re: RE : [Softwires] Analysis of Port Indexing Algorithms (draft-bsd-softwire-stateless-port-index-analysis) Hi Med, I think that exclusion in draft-murakami means exclusion from sharing. All nodes have access to ports 0-4095 plus the port range assigned. Two nodes with the same address using the FTP port, would that work? These things are very confusing :-). Regards, Behcet On Wed, Sep 21, 2011 at 12:54 AM, <mohamed.boucad...@orange.com<mailto:mohamed.boucad...@orange.com>> wrote: Hi Behcet, It is part of the service provider business to offer services which rely on the 0-1023 range even in a shared address environment (e.g., host an FTP server, etc.). Because statically not all subscribers use these features, the whole 0-1023 range may be assigned only to few subscribers (whether this is on-demand or during the service subscription, is out of scope of what we are doing here). As such, the ability to assign or not that port range should be left to each SP and not excluded by default. We analyzed two types of algorithms in draft-bsd: * algorithms which exclude by construction that range. * algorithms which leave excluding that range to the SP: i.e., the algorithm allows to generate a port range including the 0-1023 but the SP may decide to not assign it to any requesting user. This should be part of the SP policies. Cheers, Med ________________________________ De : Behcet Sarikaya [mailto:sarikaya2...@gmail.com<mailto:sarikaya2...@gmail.com>] Envoyé : mardi 20 septembre 2011 08:36 À : BOUCADAIR Mohamed OLNC/NAD/TIP Cc : Softwires-wg Objet : Re: RE : [Softwires] Analysis of Port Indexing Algorithms (draft-bsd-softwire-stateless-port-index-analysis) Hi Med, I thought excluding well known ports is good idea because all nodes need it. Below you seem to agree that sharing them may be useless. Why is it service provider issue? It seems like address sharing changes the basic characteristic of the port number space :-). Regards, Behcet On Mon, Sep 19, 2011 at 2:38 AM, <mohamed.boucad...@orange.com<mailto:mohamed.boucad...@orange.com>> wrote: Hi Behcet, Excluding 0-4095 range is not justified in those documents. Excluding by default this range may be considered by some service providers as waste of ports. Excluding 0-1023 range may be understood by "port utilisation fairness" but still be considered as an inefficiency if it is excluded by the algorithm and not let to the taste of service providers to assign or not that range. Sharing the 0-1023 between several users may be useless (e.g. is valid scenario to assign port 80 without port 443?). Cheers, Med -----Message d'origine----- De : Behcet Sarikaya [mailto:behcetsarik...@yahoo.com<mailto:behcetsarik...@yahoo.com>] Envoyé : vendredi 16 septembre 2011 17:02 À : BOUCADAIR Mohamed OLNC/NAD/TIP Cc : Softwires-wg Objet : Re: RE : [Softwires] Analysis of Port Indexing Algorithms (draft-bsd-softwire-stateless-port-index-analysis) Hi Med, Another question: On page 19, you have: The limit of 0-4095 ports appears rather arbitrary and represents a likely waste of ports, if not more that an operator may be interested in utilizing. Why? Shouldn't they be excluded? Regards, Behcet
_______________________________________________ Softwires mailing list Softwires@ietf.org https://www.ietf.org/mailman/listinfo/softwires
