Published: Sunday, Dec 5, 2010 Last modified: Friday, May 12, 2023
Packaging has interested me for awhile and of late I prefer the simple arch packaging where you just clone from a hg URL that’s importantly browsable from the Web.
Since I work with widgets, I’m often left thinking about how they compare to “the mature Debian way” of doing things.
The kill switch
This seems to be an Operator requirement, since Android and the Itunes App store I believe implements one.
On Debian and all other Linux distributions security is largely the responsibility of the user. That is to rarely manually remove or rather upgrade the package in question. It most cases an “evil” or dangerous package is detected by its maintainer or during the unstable->testing->stable transition by intrepid users, hence most removals happen in “unstable” and extremely rarely in “stable”. Security incidents are described in security advisories, which in turn are usually linked to a industry standard CVE identifier. However if a “stable” sysadmin fails to read and act upon a removal security advisory, there could be trouble ahead.
I personally like the upgrade approach as that’s very much in line of how the Web works. You update your way out of trouble, and that works for all parties.
Forced removal or disabling of an app (aka the kill switch) is a very disruptive experience. What happens if a user relies on that particular “bad” app to get his or her job done? There is no upgrade path for users.
Comparing the update process
Updates are a pain point. On Debian for example, they are done rather network inefficiently, as you need to download the entire package even though there might be a one line of change.
On mobiles, this problem worsens as network capacity is tightly limited and complex, where OTA push updates are (I’m not sure why) favoured. Furthermore Operators typically demand to test each update before deployment, making the process rather cumbersome and dangerously slow.
I personally think the manual upgrade path with efficient binary diffing is the way forward, much like the Web works.
And who does the updates ?
If a security issue is found on Debian, the security team or package maintainer usually assume responsibility for fixing the problem as quickly as possible.
In the WAC widget, Android & Apple App store world, the emphasis is squarely on the author to fix the problem or be banned/removed/killed. I would argue this is bad for users and the ecosystem at large, since continuity and stability is put at risk.
Of course it is more complicated than that. Add the whacky shrinkwrap & “freemium” business models and you generally put blame squarely on the original author. But can you? The picture is really muddled when intermediaries like Apple and alike who take a revenue share and conduct testing.
Do you think a intermediary has a right to change a product (violate “artistic” integrity) on its way to the end consumer? I’ll leave you to think about what happens in the real marketplace when you next go shopping.