On Tuesday 25 November 2008 23:31, Zero3 wrote:
> Matthew Toseland skrev:
> >> True. But don't you agree that Linux is gaining market share at the 
> >> moment, that these new users prefer the easy GUI-based distros like 
> >> Ubuntu, and the de-facto standard of installing software on these 
> >> distros are via packaging systems? Getting Freenet packaged (and 
> >> prepared to be such properly) and available in the Debian/Ubuntu 
> >> repository seems like a great step towards a more user-friendly 
> >> installation on Freenet on Linux. 
> >>     
> >
> > Including Freenet in anything other than the debian-volatile repository 
and 
> > its equivalents is grossly unrealistic at this point. And I doubt even 
> > debian-volatile would have us. We make network level behaviour changes 
> > constantly, and we need those changes to be deployed in days, not years: 
> > freezing the code short of security fixes for 3 years, or even for 12 
months, 
> > is unacceptable and infeasible. Including ourselves in the experimental 
> > repositories etc is conceivable but would have to be accompanied by a cast 
> > iron promise not to ship Freenet in stable until 1.0, and I doubt that 
debian 
> > et al would be interested on those terms.
> >   
> ... you have to work your way up from the bottom, don't you? :).
> > What we can do is provide a repository of our own, like for example Wine 
does. 
> > However, we must provide packages for at least:
> > - All debian derivatives including Ubuntu variants
> > - SuSE (not necessarily compatible with Fedora)
> > - Fedora (arguably; many geeks use Fedora and IMHO geeks are important to 
us)
> > - Gentoo (slightly weaker case than Fedora, but at least we have an ebuild 
> > already thanks to Tommy[D])
> >
> > Even then, there is a big question mark over whether the packages will be 
> > updated frequently enough. Is this a problem in practice on GUI distros? I 
> > know they generally offer a GUI update notification service, so maybe not 
if 
> > we average say one new build a week...
> > ...
> >   
> It would seem like a perfectly acceptable solutions to provide "private" 
> packages for all popular distros. 

Yes, we should have our own repository for each major linux distro family.

> Once it runs smoothly, have been  
> testing, fixed, adjusted, etc., it should be *much* easier to get 
> accepted into official repositories. I'm quite sure the official package 
> maintainers would rather move a well-tested private package into their 
> repositories than build a new one up from the ground (which also 
> requires knowledge only the devs will have). You will have to start 
> somewhere...

No point until 1.0, as explained already.
> 
> Regarding package updates: Atm. Freenet probably *does* have way too 
> many short-notice mandantory updates to rely on packaged files alone. 
> See my previous suggestion about a packaged version with "update 
> overlays" though (to sum up very short: Between package updates, Freenet 
> keeps itself up-to-date by "shadowing" the newer files over the packaged 
> ones, and when package gets updated, deletes the "shadow files"). 
> Obviously this is not a problem if Freenet is solely distributed via own 
> repositories, but if the goal is to get accepted into distro reps., this 
> needs to be taken care of.

Not going to be acceptable for distro repositories. And complicated/messy.
> 
> >> Another thing to solve is the disagreements on how Freenet should 
> >> operate on Windows. Atm. Freenet creates its own user account and 
> >> installs itself as a service, as opposed to running as a normal 
> >> background application as the logged in user.
> >
> > I'm unsure. On the one hand, creating a user account and installing a 
service 
> > is counterintuitive for Windows users, causes permissions problems and can 
> > cause the service to not be easily killable from Task Manager. On the 
other 
> > hand, on any machine with a login screen, whether or not you need a 
password, 
> > starting on startup can gain us a significant amount of uptime - not only 
> > because it will run in the background when other users are logged in, 
which 
> > IMHO is perfectly legitimate and something that the users can negotiate 
> > amongst themselves - but also because if the computer is switched on but 
not 
> > logged in, Freenet won't run at all. Uptime matters. Dealing with low 
uptimes 
> > is hard.
> >   
> Question is: Is this uptime worth it, compared to the consequences of 
> running as a service? 

Yes.

> (Wireless connections most likely won't be active  
> at the logon screen anyway, mind you, and no internet connection makes 
> it quite pointless?). 

Even fixed wireless connections, with login saved already or no login?

Wireless connections provided by cafes etc are not suitable for Freenet - they 
are likely severely throttled and severely NATed, and may bill per gig. 
Mobile computers suck for Freenet.

> Freenet will probably have to learn how to deal  
> with poor uptime with the increase of laptops and mobile internet 
> connections anyway... 

We do, but it's a *hard* problem that we can't deal with right now.

> I don't have any stats at hand, but I'd argue that  
> wired machines on logon screens are responsible for a very little part 
> of the general Freenet uptime available. And it will most likely only 
> get smaller with time anyway...

It doesn't matter whether it's wired or not, it matters whether it's *fixed*. 
Wireless can still be fixed.

> > If we install as the logged in user, we don't need to create an account, 
we 
> > can simply create a user-specific startup item. This would certainly 
simplify 
> > installation.
> > Maybe we should ask the user? But that means more steps in the 
installation 
> > process!!
> >   
> If we (as previously discussed) move the autorun question to the wizard 
> and separate the identity from the program files, we can simply add 
> Freenet to each user's startup group as we are given permission to from 
> the users (assuming every user will get to complete their own copy of 
> the first-time wizard). Maximum userfriendliness...

No, we can't. It's a web interface, the user cannot grant us anything, short 
of typing in his password to a form, which IMHO is baaaaaaad.

Also, if the user does not immediately complete the wizard on install, and we 
don't auto-start, will it be running when he wants to use it? I guess the 
answer there is for Browse Freenet to include Start Freenet.
> 
> >> Machine-specific settings: "/etc/freenet"
> >> User-specific data: "~/.freenet"
> >>     
> >
> > IMHO there is no user-specific data for Freenet. It's a daemon, like 
Apache. 
> > It may implement its own user mechanism, 
> >   
> The identity? Friends (darknet)? fproxy theme and similar settings? All 
> seems user-specific to me?

None of it is per user. It's a web interface, remember? We have no idea which 
user is logged in; if we implement a login mechanism in future, we will have 
to maintain our own username:password database.
> 
> - Zero3
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 827 bytes
Desc: not available
URL: 
<https://emu.freenetproject.org/pipermail/devl/attachments/20081126/1117cbac/attachment.pgp>

Reply via email to