[gnso-rds-pdp-wg] Five models of RDS (was Re: Apologies, and some reflections on requirements)

[Mark Svancarek]

Yes, I should clarify my statement, because I contradicted myself a bit.

I really only like Model V if someone I trust is running it :)  

That's why I mentioned my concern about ICANN SLA.  I don't think they are ready to run a 99.999% online service.

Sorry for the confusion.

-----Original Message-----
From: gnso-rds-pdp-wg-bounces at icann.org [mailto:gnso-rds-pdp-wg-bounces at icann.org] On Behalf Of Andrew Sullivan
Sent: Thursday, June 30, 2016 11:03 PM
To: gnso-rds-pdp-wg at icann.org
Subject: Re: [gnso-rds-pdp-wg] Five models of RDS (was Re: Apologies, and some reflections on requirements)

On Thu, Jun 30, 2016 at 07:32:45PM +0000, Mark Svancarek wrote:

> Model V is the one I'd build if I weren't so concerned about the plethora of local privacy laws and law enforcement regimes.

There is another thing about Model V I didn't point out but that I think is worth noting.

Model V is monolithic in that anyone on the whole Internet who wants to look at anything out of the RDS has to contact this single service.

Everything we know about how the Internet has scaled well suggests that monolithic services are extremely hard to do well.  The things that have really gotten huge are of two types:

    1.  Distributed systems that are mostly cheap to operate.  Think
    DNS, the web, and so on.  Certain large operators have an
    expensive installation, but no individual service is super
    expensive to operate and if it fails it doesn't take down the
    class of service completely.

    2.  Massive single-company category killers that depend on
    advertising revenue, revenue gained by knowing a lot about users
    and selling that, money dependent on a "magic happens here" belief
    on the part of investors, or paid use (or all of these).  Think
    Google, Facebook, Twitter, Office360, and Amazon (both the
    commerce site and AWS) -- or maybe pets.com for the third category
    of these.  Notable here is that if the operator has a bad day the
    entire _class_ of service disappears.  There is no alternative
    Facebook: if they're broken, Facebook stops.  (Fortunately,
    they're very, very good and rarely have this happen; but that's
    not an operation built on a shoestring.)

The plan for a monolithic RDS is basically to build (2) and hope that revenues and operations staff adequate to (1) will be enough.  I hope it is self-evident what the problem is here.  Moreover, I hope that everyone involved in this WG is familiar enough with the term "DDoS"
to see why building a Big Giant Centralised Service might be like painting a target on ICANN.  Or perhaps _another_ target.

Best regards,

A

--
Andrew Sullivan
ajs at anvilwalrusden.com
_______________________________________________
gnso-rds-pdp-wg mailing list
gnso-rds-pdp-wg at icann.org
https://na01.safelinks.protection.outlook.com/?url=https%3a%2f%2fmm.icann.org%2fmailman%2flistinfo%2fgnso-rds-pdp-wg&data=01%7c01%7cmarksv%40microsoft.com%7c5add48b130a044d1574b08d3a1219cee%7c72f988bf86f141af91ab2d7cd011db47%7c1&sdata=4p0L3a%2bOw4wi8Wqjl7SU5Y6vKGrX1w3yK1JmoIzgFRI%3d