netdev
[Top] [All Lists]

Re: [PATCH] anycast support for IPv6, updated to 2.5.44

To: davem@xxxxxxxxxx
Subject: Re: [PATCH] anycast support for IPv6, updated to 2.5.44
From: YOSHIFUJI Hideaki / 吉藤英明 <yoshfuji@xxxxxxxxxx>
Date: Thu, 20 Mar 2003 12:44:28 +0900 (JST)
Cc: dlstevens@xxxxxxxxxx, kuznet@xxxxxxxxxxxxx, linux-kernel@xxxxxxxxxxxxxxx, netdev@xxxxxxxxxxx
In-reply-to: <20030319.192331.95884882.davem@xxxxxxxxxx>
References: <20030319.163105.44963500.davem@xxxxxxxxxx> <20030320.120136.108400165.yoshfuji@xxxxxxxxxx> <20030319.192331.95884882.davem@xxxxxxxxxx>
Sender: netdev-bounce@xxxxxxxxxxx
In article <20030319.192331.95884882.davem@xxxxxxxxxx> (at Wed, 19 Mar 2003 
19:23:31 -0800 (PST)), "David S. Miller" <davem@xxxxxxxxxx> says:

>    > I'm going to apply this, with the small change that dev_getany() is
>    > renamed to dev_get_by_flags() which more accurately describes
>    > what the routine does.
>    
>    Again: I don't like API at all.
>    
>    Anycast address management itself in that patch would be ok.
>    However, JOIN/LEAVE is NOT useful and userland application will be 
>    incompatible with other implementation. (sigh...)
>    I think linux likes unicast model (assign address like unicast address), 
> too.
>    
> Please propose alternative API, or do you suggest not
> to export this facility to user at all?

I like to assign address like unicast (using ioctl and rtnetlink 
(RTN_ANYCAST)).
We suggest you not exporting this facilicy until finishing new API
(And, another API would be standardized;
This is another reason why I am against exporting that API for now.)

-- 
Hideaki YOSHIFUJI @ USAGI Project <yoshfuji@xxxxxxxxxxxxxx>
GPG FP: 9022 65EB 1ECF 3AD1 0BDF  80D8 4807 F894 E062 0EEA

<Prev in Thread] Current Thread [Next in Thread>