[Dev] Multiuser : Tizen Login Manager : should it be optional

Dominig ar Foll dominig.arfoll at fridu.net
Thu Oct 31 14:38:35 GMT 2013


following discussion on this mailing list and in 1:1 including Linux Con
EU, last week, in Edinburgh which was a great place to chat face to
face, I arrive to the conclusion that making the Tizen Login Manager
(TML) optional may not be the smartestway to enable support of single
user device in Tizen 3.0.

1) Introduction
Tizen 3.0 will support multiuser mode but it does not mean that all
devices will need that extension. Avoiding to add unneeded extra burden
on small devices remains a clear architectural focus.

2) TML
TML concept regroups a set of utilities and libs which allow to add user
and remove users as well as to log a user on the fly.
TML does NOT provide a UI which means that the login can be done
silently at boot time if desired.
Single user devices can be initiated using TML by login a default user
at boot time. In that type of use case, adding the utilities asociated
with user management in the image is not needed.

3) Unified architecture or fork
Imposing TML in the base Tizen 3.0 allows to unify the architecture for
single and multiple user devices.
Making TML optional would allow to optimise further ressources on single
user devices at the cost of an architure fork which will likely be paid
at the high price when it comes to code test and maintenance.

4) Compromise
As TML does not require any UI, single user devices can offer the same
service to the user with or without TLM.
My proposition is to impose to the TML team to create the TML service in
a modular way, allowing for a minimalistic implementation when applicable.
E.g. there is no need to drag FOB and login UI when that i not required.

If such constrain is enforced in the initial TML design, we could leave
the TML as mandatory for Tizen 3 base, avoiding a fork, and adding only
a few KB of extra code in single user devices.

Comparing the cost of maintaining a fork versus a few kB of RAM, I feel
that a fork is an extrem model that is not justified in this case.

So I propose to remove the optional status for TML and its associated
services in the multiuser architecture release candidate.

Your feedback will be appreciated.


Dominig ar Foll
Senior Software Architect
Intel Open Source Technology Centre

P.S. Tizen gouvernance clearly states that if you do not speak up, you de facto, agree. So if you agree, there is no need to respond :-)

More information about the Dev mailing list