Mohit Singh writes: >> Just by accumulating books you won't become a wizard, otherwise all >> librarians >> become Faradays :). You've to study them. The network they provisioned to me >> was half duplex and on a broadcast domain. I used to run IPv6 radvd in my >> network, and I announced it on their interface and within minutes everyone is >> showing their ICMPv6 love to me :). I can tcpdump and I see everyone's >> traffic >> on my interface. Looks like they put it into a hub. >> >> And then there is Airtel, who has it's own tale ;). So it is not about >> accumulating devices but knowing that best configuration for the job.
> You are missing a crucial point here. Every ISP has to combine a
> number of devices from a number of vendors who dont want to see the
> faces of the competition.
The key is not here accumulating n devices of m vendors, but making sure
they're configured properly.
> If they rely on one vendor, they'll become slaves of it. So, they have
> to build a multi-vendor architecture practically.
> 'Interoperability' is biggest challenge among all those devices. I
> remember how Alcatel was showing us their end to end solutions for
> almost anything back to back with its own core and edge devices. when
> we asked them to connect to mini replica of our core and metro rings,
> they JUST REFUSED. They said they cant guarantee performance if their
> devices interface with others in the core or access rings :) And this
> is just one instance.
This happens with proprietary vendors, badly written code, not worth making
public, worst with interoperability.
> Every ISP has some problem with some part of their network. Reliance
> has one of the fastest core. They even had a successful IPTV project
> way back in 2001 :) that too with the legendary Mohan Tambe ( I hope
> you know him). That does not mean that OSS segment may not work badly.
> Yes, there were problems with Nortel Clarify and even 8800 boxes.
Yes, they can solve the problems, but instead they live with it as long as
people don't complain. Customers on residential DSLs don't get any SLA
guarantee but that doesn't mean they don't deserve quality. Oh, and running
things in lab is different from running it in production. I never heard of
Mohan Tambe. And I hope you can see the progress with IPv6 which is probably
running in (test?)-beds of Indian ISPs since years, but yet to see a
successful deployment.
> All because interoperability test reports of these vendors is nothing
> but a bunch of nonsense.
>> Could be, but they chose Linux. I reported no. of bugs with the IPv6 code. I
>> think the problem with BSDs, is they've less and overloaded-with-work
>> developers and thus less eyeballs...
>>
>> Linux kernel has probably more contributors than BSDs, and more sponsors.
>> Most
>> of the people work with FreeBSD do it in there free time, though there are
>> some lucky ones who get paid to work on FreeBSD :)
> I asked Ibrahim Haddad - the leader of 'Carrier Grade Linux' project -
> Do you intend to give competition to JUNOS? He said - we intend to
> give competition to Solaris :)
> Who is paying Alexey Kuznetsov for doing IP stack in Linux Kernel?
> What is the use of EPFL doing RSVP in Linux when no ISP can use a
> Linux code for pure 'end-to-end' thing. More chances are there for
> having BSD/BSD derived codes working in their core and edge devices.
> BSD TCP/IP got parents like Bill Joy, Von Jacobson, Sally Floyd and
> many more. All of their research speaks of nothing but BSD based
> systems. TCP/IP was born in BSD and even benchmarked in BSD.
This is completely FUD. I hope you know that TCP/IP was born in BSD because it
existed during that time, source were freely available. There are RFCs,
working groups, and things which don't get standardized, were written to be
compatible with an already existing implementation. And it's purely
implementor's choice. {S,}he might have some technical reasons to go for one
and not other but not always.
> The documentation of Linux IP does not have even the 'discussions' which
> formed the basis of their TCP/IP.
And why is documentation going to contain the archives of discussion, that
doesn't make any sense.
> What we find are direct coder's views at LKML and all those point to
> ultimately the things done in the BSD world.
BSD is a popular implementation. BSD sockets API is a popular API and works
great. Why would anyone implementing an OS wouldn't implement BSD sockets API
just because they want to look different. This argument can work, but Linus'
goals was to build a UNIX like system on a cheap 386 boxen.
>> Right, but you started with blind accusations.
> If you wish, I'll come up every time with example now. Though they
> were not blind, but a gist of experiences.
>> freebsd.org!ashish | http://people.freebsd.org/~ashish/
> why just 3 lines there :)
> You have not mentioned about your area of work. I could learn more from you.
It'll soon have something.
--
Ashish SHUKLA | GPG: F682 CDCC 39DC 0FEA E116 20B6 C746 CFA9 E74F A4B0
freebsd.org!ashish | http://people.freebsd.org/~ashish/
Avoid Success At All Costs !!
pgp8AGUTni6Ic.pgp
Description: PGP signature
_______________________________________________ bsd-india mailing list bsd-india@bsd-india.org http://www.bsd-india.org/mailman/listinfo/bsd-india
