| 1 |
On Wed, Dec 19, 2012 at 4:26 AM, Alan McKinnon <alan.mckinnon@×××××.com> wrote: |
| 2 |
> On Tue, 18 Dec 2012 17:55:16 -0500 |
| 3 |
> Michael Mol <mikemol@×××××.com> wrote: |
| 4 |
> |
| 5 |
>> On Tue, Dec 18, 2012 at 9:33 AM, Alan McKinnon |
| 6 |
>> <alan.mckinnon@×××××.com> wrote: |
| 7 |
>> > On Tue, 18 Dec 2012 09:08:53 -0500 |
| 8 |
>> > Michael Mol <mikemol@×××××.com> wrote: |
| 9 |
|
| 10 |
[snip] |
| 11 |
|
| 12 |
>> > #2 already has a solution, it's called an init*. Other solutions |
| 13 |
>> > exist but none are as elegant as a throwaway temporary filesystem |
| 14 |
>> > in RAM. |
| 15 |
>> |
| 16 |
>> I find virtually nothing elegant about a temporary filesystem in RAM. |
| 17 |
>> It duplicates code that already exists on the system, and it |
| 18 |
>> represents and additional maintenance step in system upgrades. It |
| 19 |
>> seems almost a given that if someone is keeping multiple kernel images |
| 20 |
>> on a system, they're not updating the initr* for each when binaries |
| 21 |
>> that would be found in each are upgraded or rebuilt. |
| 22 |
> |
| 23 |
> You could level the same criticism at boot loaders... |
| 24 |
> |
| 25 |
> The also duplicate functionality in the main system they launch (albeit |
| 26 |
> read-only at least in legacy grub) in the fs drivers, are only used |
| 27 |
> briefly at boot-time and have to be maintained. True, the bootloader |
| 28 |
> doesn't contain a *copy* of the main system files which is where the |
| 29 |
> parallel breaks down. |
| 30 |
|
| 31 |
Bootloaders tend to be updated in a targeted or atomic fashion. This |
| 32 |
remains true even as they become more and more like full operating |
| 33 |
systems. |
| 34 |
|
| 35 |
> |
| 36 |
> init* may well suck, but they suck less than any other possible |
| 37 |
> solution. |
| 38 |
|
| 39 |
I disagree. I think defining boot stages, and fixing cross-stage |
| 40 |
dependency errors, is a far better solution. |
| 41 |
|
| 42 |
> And they come with one more benefit - the community has |
| 43 |
> agreed on how they work so they form a standard of sorts |
| 44 |
> |
| 45 |
>> |
| 46 |
>> In Debian, Ubuntu and others, this is handled by a post-install hook |
| 47 |
>> where the initr* image is rebuilt. To me, this honestly feels like a |
| 48 |
>> hack. In something like Gentoo, I'd rather see package placement |
| 49 |
>> driven by whether or not it will be needed to get all mount points |
| 50 |
>> mounted. If that means i18n databases under something like /boot/data, |
| 51 |
>> that seems reasonable. To me, the only cases where initr* feels like |
| 52 |
>> the right solution are things like netboot or booting from read-only |
| 53 |
>> media. |
| 54 |
> |
| 55 |
> Let's not forget the prime reason why init* exists - so that binary |
| 56 |
> distros can boot and still have all kernel modules required at launch |
| 57 |
> time compiled as modules. |
| 58 |
|
| 59 |
Another valid use case. |
| 60 |
|
| 61 |
> |
| 62 |
> Bootstrap code and operation is ugly, always has been and always will |
| 63 |
> be. It also tends to be inflexible unless you use a slipstreaming |
| 64 |
> technique (my invented description of how init* works). I still think |
| 65 |
> the solution is elegant given the real-world requirements imposed on |
| 66 |
> systems. |
| 67 |
|
| 68 |
And I still think that, outside of scenarios where workarounds don't |
| 69 |
exist, it's an ugly hack and a cop-out. |
| 70 |
|
| 71 |
-- |
| 72 |
:wq |
Archives