Team-BHP - The Mobile Phone Thread - Queries, decisions, discussions all here

[QUOTESlowly by late 60s it was stabilized and used on many big machines][/quote]It was, surely, stable when it was used one AT&T text-processing system?

Quote:

Why the past tense here? You better edit before time runs out. Just joking.

:rofl:

Certainly not all unixes were/are equal. i worked with RS/6000s and AIX was not popular with everybody, but it was a great improvement on IBM's previous Unix incarnation: people said they were lucky to keep the thing running for a day!

Multics was abandoned. it is not really part of the Unix history, just something that left people with nothing much to do, and the source of the joke that unix was, originally, not multi-user.

Anyway... it was all a long time ago now! :)

Quote:

Why the past tense here? You better edit before time runs out. Just joking.

Well, it was in the historical context :)

It is a well-known thumb rule within product development communities. Any core product development team should not be bigger than 5 people. I have found this to be true as I have worked in both large and small companies as product developer. Any bigger, they will be spending more time in meetings than coding.

Any large OS like Windows or Linux is not a single entity. There so many core teams within that product. Each of them will be small. There is no other way to manage it.

That said, there are many companies that are dumb enough not to follow this. Usually happens when MBA types try to control core development.

Having worked in AT&T Bell labs, I can give some insight here. Until 1984, when they had totally monopoly, Bell Labs created some of the greatest products whose benefits we still reap. Bell Labs never really made any money out of Unix, C, C++, korn-shell, etc. But after the 1984 Ma Bell breakup, MBA types came in and tried to control research and product development. Some of the biggest product disasters happened during this period. The NET 1000 project was cancelled after they spent nearly a billion dollar on it. Imagine, a billion dollar in the 80s, on one project. I was surrounded by NET 1000 survivors when I worked there. They used to imagine fantastic products, throw 100-200 engineers into it and then cancel it after a year. I was in two such failed projects there. Managers spent all their time in meetings, and most of us in the team couldn't have explained the product even at gun-point. We only knew our nuts and bolts. Such was the disaster of large size product teams without clear focus. The year I left, they further broke down to Lucent and many other companies. Few years later they again Lucent broke into Avaya and others. Biggest lesson I learned there is the importance of small teams with clear focus.

Quote:

Originally Posted by Samurai (Post 2247078)

Any large OS like Windows or Linux is not a single entity. There so many core teams within that product. Each of them will be small. There is no other way to manage it.

Exactly thats from where this discussion started any mobile OS today has dozens of subsystem in App Platform and this is collectively called OS not just kernel. In fact Linux Kernel being monolithic itself has dozens of systems contained in it ranging from FS to ALSA to TCP/IP requiring different domain skills and taking ball park figure of 5 developers in each system the overall becomes huge. As kernel is not from scratch in case of mobiles emphasis is on middle ware and driver port of this kernel.

So taking 50 -75 middle ware systems the core team size itself grows to 200-400 for development. And this is as close to real figure as I can go.
Take example of Android itself Andromeda started the project in early part of this decade and gave demo.
Google bought them over in 2005 and took 2 years to finish and release first version SDK.

Since Android is in open source it is easy to refer to systems / subsystems but on the platforms I have worked the numbers remain more or less same. So all that single small team assumption do not hold true for modern OS be it Linux /Android/WP7/Symbian or any other.

Quote:

Originally Posted by BaCkSeAtDrIVeR (Post 2247223)

Ahh. Now I get the point.

Now I am getting curious, what does android use in place of xorg?

This is biggest weekness of Android as of GB and also Honeycomb. It does not have a client-server arch of X that is removed instead it uses a UI layer which has widgets ( equivalent of GTK /QT) written in Java and the Graphics system is provided by SKIA so you can not have multiple independent UI clients like X clients. So on Android phones / tablet you have full multitasking but the UI of 2 apps can not run side by side lets say on Half screen one App and another app on another half like X based systems. You can dig out my old posts on Android thread for this :D
There is a solution being developed somewhere, in-fact already working you will get to see only if it gets accepted

Also There is a QT port available on Android but that is using OpenGL-ES back-end for rendering and not the usual X arch.

Quote:

Originally Posted by NetfreakBombay (Post 2246748)

I agree with that. Smaller teams working in close proximity find it easier.

Quote:

Originally Posted by amitk26 (Post 2247007)

The basic premise of Navin's assertion was that OS written by small team are stable. The stable industrial Unix was not a small one man or small team project at all.

Every prototype starts as a one or two man project and at that point when it is programming in small it remains largely unusable as the Unix was when Ken Thompsan started on it in first avatar Multics on PDP7 mini computer.

The AT&T team was still small compared to the team MS dedicated to write Windows.

BTW I was refering to "Core" teams not teams for shells, applications, and other services that sit on top of the core.

If the core team is small and produces a tight stable core you can build any number of flavours, services, applications or shells on that core which will in varying degrees inherit someof the stability of the core.

BTW I may not be PC literate today but in the 80s I was. I used Unix on a PDP 11/70 and later on a variety of IBM machines and Covergent (PC7300 or PC3700 I think it was called) workstations. Convergent was later acquired by Unisys (aka Sperry-Burroughs).

Quote:

Originally Posted by mercedised (Post 2247242)

"I want a phone under 15,000 Rs which has good GPS and WIFI capabilities and a good camera too. I donot need a touchscreen (a partial or even no touchscreen phone) nor do I need a Business phone."

If you are not particular about Qwerty keypad, then I suggest Nokia E52. It has 3G, GPS & Wifi. Its been year since it was launched here. You might get for 10K. It has mighty battery life. Camera is 3.2 MP.

Quote:

Originally Posted by navin (Post 2248097)

If the core team is small and produces a tight stable core you can build any number of flavours, services, applications or shells on that core which will in varying degrees inherit someof the stability of the core.

Navin Ji we are talking of a consumer elctronics device here , Any allegory of mainframe / mini computer or even Unix workstation era is not applicable becuase users then were basically engineers /technical people or were backed by a system admin.
A mobile phone, gaming device, Camera or TV user is not expected to deal with core OS or use simple shells neither he is going to sign up to a sys-admin service.

Today as well the Core OS bring-up on target device happens in around one month by a single engineer if hardware is new , Annother couple of weeks by a small team to write all the drivers and bring up peripherals but armies are required for everything else for year or so. As there is not much change in core OS ( kernel) which is always essentially developed by a small team no effort is put to fiddle with it.

What is known as OS today is Kernel + UI Shells + middleware + tools+ basic applications else Android / Meego / Maemo / LiMo /MOAP-L are exactly same as far as core goes ( yeah just few differences in configuration )

Quote:

Originally Posted by navin (Post 2248097)

BTW I may not be PC literate today but in the 80s I was. I used Unix on a PDP 11/70 and later on a variety of IBM machines and Covergent (PC7300 or PC3700 I think it was called) workstations. Convergent was later acquired by Unisys (aka Sperry-Burroughs).

Well so in 80s you were definitely not using a one man Ken Thompson version of Unix it was definitely matured by full 1.5 decade of usage and development by various people.

BTW my second computer was a PDP 11 , no I did not study in that era but just that antique peace was not discarded in our not so modern department and was used to teach FORTRAN programming :D till early 90s so I can say it was very very evolved Unix.

Quote:

Well so in 80s you were definitely not using a one man Ken Thompson version of Unix it was definitely matured by full 1.5 decade of usage and development by various people.

My "era" too. What you say is true: by this time, and by the time most of us commercial users came to Unix it had already done plenty of too-ing and fro-ing and been through many hands. I remember that even I reported a bug in a basic utility ("ls not display symbolic links") in the Unisys/NCR box implementation.

Historically, though, in the early days, there was a methodology of development and documentation that was very different to chuck-it-out-the-door ethos that Microsoft perfected.