Pleng said:
I certainly never want to be in the position where one SD always HAS to be inserted (presumably in the same SD slot). I'm often ejecting SDs to download applications onto them with my laptop (much faster wifi) and as time goes on I imagine I'll be augmenting my current SDs with a couple of 32Gig cards. My 16 and 8 gig cards will then be used for films and music. Again I don't want to have to keep an SD card in place.
I never said you have to keep an SD card in place. I'm just suggesting that the default behavior should be that if no card is installed you can monkey around with your system all you want, and everything gets set back to default if you reboot. It's not that terrible. It protects your investment. It's good from a developer's standpoint because they don't have to worry as much about writing to home and it significantly reduces the time needed to port a large number of useful applications.
I know Skeezix once said that the developer can just "export $HOME" to somewhere else. Unfortunately for us, Gnome applications most often use g_get_home_dir() to locate home, and it will use the REAL home. I guess it's a security thing. I don't know, but that function bypasses the "export" trick. So instead of a simple grab the source and recompile for most of the applications in the Angstrom Repo. Devs will have to sniff out these g_get_home_dir() calls and check to make sure it's safe to replace them with g_get_current_dir() in addition to doing the export $home trick. It's not ideal, and eventually porters will care less about writing stuff to /home/[username] and things will get increasingly messy for the user. In the case if Remmina, even though I wanted to make the program never touch NAND, I couldn't because it looks for /home/[username]/.ssh which is managed outside of Remmina by the already installed SSH client. I couldn't help but write to your NAND. This is also not ideal.
Besides, you already have the option of having HOME on an SD card; just BOOT from an SD card and HOME will be on that card. Don't impose your lifestyle choices on the rest of us though!!!
Right! Sinoth's tarball works pretty good, but it has a few small issues for me, and it's not an official solution. Setting up a card to boot from is no picnic, especially for Windows users who can't easily format a card to EXT2 or extract the files with permissions intact. But I still opt to this. However, it's not really about lifestyle choice. It's about what's good for everyone in the long-term. Your Pandora would still perfectly usable without an SD card, and there's probably no reason why /home/[username] protection couldn't be toggled on and off so people could choose. What I'm saying is, it should be an EASY option for end users to take. They shouldn't have to jump through hoops to protect what should be protected in the first place.
In a perfect world, no program, except for firmware updates, should ever write-back to the NAND, unless absolutely neccesary. As long as home is on the NAND, everything you do is touching NAND. Open a terminal, .bash_history is getting written to every time you type a command. Using Evolution to check your mail? Guess where that's getting downloaded to? Using Fennec, Firefox, Midori, Arora? Guess where all that browsing history is getting written to? Open a PND that wasn't groomed to be NAND friendly, all that crap goes into your home folder, out of sight - out of mind. It's not in the /pandora/appdata/[pnd_id] folder where you expect it. It just took a whiz on your NAND for no good reason. The NAND is very limited on space and with each patch that space is only getting tighter. It's only a matter of time before this becomes a real problem for some people.
@WizardStan, while that is a "servicable" solution it:
1) is only good for people who are aware that this is a problem with the current setup to begin with
2) is only good for people brave enough to make those changes on their own
3) doesn't alleviate the burden of fixing the g_get_home_dir() issues in Gnome applications from the devs.
I really hope that the OP Team recognizes this as a very real problem for end users and developers alike, and doesn't continue to look at it as a matter of convenience.