17 March 2009

I saw mommy, kissing Santa ...

Santa and mommy
I can see her lying back in a satin dress
In a room where you do what you don't confess
I could picture every move that a man could make
Getting lost in her loving is your first mistake

   -- Sundown, Gordon Lightfoot

It is always kind of a sad moment, watching a younger idealist encounter something that tears asunder their old mental model, and puts them on the path to being a battered, old, steel eyed mercenary. But with that loss of innocence, new doors open

One useful paradigm to look at the consumers of Enterprise *nix software is to break them into a partition of three major types:
  1. Those that Have to have the 'Real McCoy', possibly for 'CYA' purposes, or because a upstream vendor says that they need the 'real' one as part of the 'silo' they will support without extra charge (if at all) to meet a performance SLA
  2. Those who do not have a strong mandate, but are generally willing to pay the minimal incremental cost such a subscription adds to their bundle of functions, and
  3. Those who will simply not pay for 'free' software: No how, no way; no, sir
The commercial enterprise Linux' have been generally successful in 'cannibal conversions' of enterprise consumers of 'olde skoole' proprietary Unix -- The morning's news has rumor that IBM is sniffing around JAVA. We covered the topic, and Ted T'so's proto-quant thought piece on this [Ted being on leave of absence from IBM to the Linux Foundation, as I recall] some months back, in the context of the future for software freedom

All the young idealists from the BSD side of the FOSS house saw their holdings of SUNW eroded away in recent years, the progressive shifts away from hardware, away from ksh v. csh language debates, into the tangled place of license issue and re-inventions of approaches on scaling, as their firm flailed with JAVA [v. rather than use the one true type safe modern OO language, c++ // sorry, could not resist], into databases with a product that will NEVER be Oracle DB, no matter how hard it tries

JAVA felt it had to move past Berkeley DB, and darn it, all the cool kids use SQL. ORCL is the only credible lead player in database space (IBM and DB2 are there of course, but databases are rounding error to IBM's financial statement). JAVA never could articulate the unique value proposition that picking up MySQL, AB, brought to the table, and let the acquisition languish, perhaps hoping that the database's engine in the 'LAMP' stack would pull in tier 2 conversion sales (see the next part, infra). I think they have pretty well demonstrated that "hope" is not a business strategy to follow

Then there is that second tier -- FOSS *nix in through the side door, and without formal support contracts at first. "Under the Radar", so to speak. [Note: The linked article is a bit 'snarky' about Bob's new venture: Lulu, but I find it a wonderful and reliable service, to convert 'print pre-flighted' PDF's to bound books, for cheap, fast and reliably. Highly recommended.]

Just as I might choose to burn up a laser printer to print a manual, and do home-brew binding, Lulu has found a value proposition that makes me 'buy' their service, rather than 'build' it myself. They have convinced me that outsourcing my printing to tier 1 is the 'right' decision. He has converted me to producing wonderful documents from TeX that his business handles the ink to paper, binding and delivery parts. It seems Bob is also 'whiteboxing' short run, 'just in time' print of conference manuals, and continuing education materials. A nice niche, but low barriers to entry

And then there are the 'No how, no way' school in tier three. This recent post in the CentOS forums, "leasing CentOS5 from DataCenter", caught my eye:
Recently we had a customer come to us asking how much we lease out CentOS for.
I thought this was an odd question - since CentOS is ... FREE

When in dialogue with them I learned they have a number of servers with a different provider that charges them $5.00 per month for the Operating system.

I thought this was a bit strange - and wondered - Is it even legal?

How can a datacenter lease out something that is free?
I could understand perhaps charging a setup fee based upon a customers requirements - this is a service --- but
for a datacenter to live off of the backs of someone else by charging for something that is free -

it just bugs me and rubbed me the wrong way -

Any thoughts - ?

Not sure why it bugged me so much - perhaps its because we write a ton of opensource software and could not imagine someone charging for the software itself.

Support / Installation / Service yes - but the software ... i thought thats what GPL protected folks from
This poster has missed the point of the GPL so widely, it is painful.

The GPL is perfectly fine with charging for software which requires that it be accompanied with an offer of access to the sources it was built from. This is what builds markets, and indeed, what makes CentOS possible in part. CentOS is fine with a redistribution and commercialization, so long as our marks and brand are not mis-represented. [Advert: The CentOS project would put a 'tithe' of that rental to good use -- money, machines, bandwidth, and so forth, but it is not mandatory.] A better question might be: Is the data center that employs that poster itself providing the GPL required offer of sources access, and meeting its duty to provide, when they provide binaries under 'lease'?

Someone may well come along and undercut a person selling GPL and related FOSS licensed software for less. I wrote a post encouraging people who 'cannot wait' for the CentOS 5.3 respin, or the updates which get stacked up, waiting for that stabilization process to end, encouraging them to 'outcompete' CentOS. I am fine with that. I know it won't happen generally [Scientific Linux is the closest credible 'fellow traveller' remaining on this highway; Hi, Connie and Troy] soon, as it is non-trivial to ship and support the full line product

The protection of the Four Freedoms under the GPL makes it inevitable that someone will make a run at commercializing FOSS; this is a 'Good Thing'. But then the trick is to provide value; that is, also provide design services, consulting, 'service after the sale', or build a support infrastructure, to make it safe to entrust one's most valuable assets to that software. I feel CentOS meets that test in the 95% case for tier 2; others may dial that number up or down, and do according to their risk tolerance

And with that, we are back to my post sending people with an external factor 'beating on them' about SLA's, to: Go Buy from CentOS' Upstream

Disclaimer: I hold direct positions in JAVA (minimal to get keep skin in the game, and to remind me to follow it) and ORCL, and have held IBM in my past; I regularly quote against IBM as to providing third-party *nix support services. I probably need to write a Caveats and Disclaimers post