Fedora Considers Dropping Delta RPMs

[gbcox]

I believe this article misses the point of Matthew's post. It isn't that DeltaRPMs as an idea is useless, or that they aren't needed anymore. We still haven't reached the point in the world or for that matter in the USA that high bandwidth internet connections are ubiquitous and inexpensive. In theory, DeltaRPMs address this issue and offers real value. The problem is the Fedora implementation has issues, and Fedora has limited resources and has yet to fix those issues - so the question is are there other things that Fedora should spend it's limited resources on that provide greater value to the user community.

[arun54321]

When android can use delta updates fine (both for OS updates and user app updates) why fedora could not implement it properly? I think it is a skill issue of fedora maintainers.

[cb88]

I believe this article misses the point of Matthew's post. It isn't that DeltaRPMs as an idea is useless, or that they aren't needed anymore. We still haven't reached the point in the world or for that matter in the USA that high bandwidth internet connections are ubiquitous and inexpensive. In theory, DeltaRPMs address this issue and offer real value. The problem is the Fedora implementation has issues, and Fedora has limited resources - so the question is really are there other things that Fedora should spend it's limited resources on that provide greater value to the user community.

That's definitely true, 7% of the US doesn't have acess to broadband at all and 42million people don't have broadband. This gets even more stark when you realise that the FCC still defines 1.5Mbit ADSL as "broadband" ... when the fact is a 600MB update would take over an hour on ADSL at that speed (and maybe much much longer depending on the quality of the connection).

About 20 years ago a the lawyer that lived in my parents house sued TWC for overcharging him, and they blacklisted the address... my parents bought the house and now have 3Mbit DSL that in practice is only 2-2.5Mbit. And that is the only terrestrial service available.

[gbcox]

When android can use delta updates fine (both for OS updates and user app updates) why fedora could not implement it properly? I think it is a skill issue of fedora maintainers.

It isn't a skill issue with the package maintainers. It is an issue with the internal build system. Go to the mailing list, read the initial post and the associated pagure ticket.

[gbcox]

I am not familiar with the deeper technological issues here, but I think it's important not to assume everyone has very fast Internet. A few months ago, due to financial difficulties, the only Internet I could afford was a shoddy 3M/sec ADSL connection. It's very frustrating when most sites/distribution systems assume everyone has good Internet. Linux is also very useful in lower income places where Internet may not be as good.

I recognize that there were other technical considerations, and I can't speak to that. If the system wasn't working the way it was supposed to, that's a consideration. I do think we ought not to assume that everyone has good Internet, because not everyone does.

100%. This article implies that DeltaRPMs are no longer useful because of ubiquitous broadband. That isn't correct. The problem is the Fedora implementation of DeltaRPMs has issues which aren't easy to fix and would require quite a bit of resources to resolve. The decision is whether to fish or cut bait.

[NSLW]

That's definitely true, 7% of the US doesn't have acess to broadband at all and 42million people don't have broadband. This gets even more stark when you realise that the FCC still defines 1.5Mbit ADSL as "broadband" ... when the fact is a 600MB update would take over an hour on ADSL at that speed (and maybe much much longer depending on the quality of the connection).

About 20 years ago a the lawyer that lived in my parents house sued TWC for overcharging him, and they blacklisted the address... my parents bought the house and now have 3Mbit DSL that in practice is only 2-2.5Mbit. And that is the only terrestrial service available.

What year are your data about FCC defining broaband from? According to an article at washingtonpost it's 25 MBit DL and 3 MBit UP.

As for the DeltaRPMs in Fedora then it could be utilized more. My gains on downloaded data through dnf update are often reported to be at 1 to 2%. Even if everyone in the world would have high speed internet connection then it's better CO2 wise to limit the data downloaded by people if it can be paired with maintaining the same level of usefulness for them.

[Vistaus]

fedora's implementation was always half-retarded.

downloading 60mb of info just to maybe save 2.5mb!

suse's (leap as TW, sadly, doesnt support it) is way better

TW supports it just fine. Just look in /etc/zypp/zypp.conf

[Vistaus]

Google chrome & derivates should be the ones vocal about binary delta updates ... just imagine how much CO2 would be saved if the whole world would not need to download 100mb+ every other day when chrome decides it's outdated ...

Except that it doesn't save CO2 at all as the CPU has to work harder. From openSUSE:

“Using a delta rpm will decrease the download size for package updates
since it does not contain all files of the package but only the binary
diff of changed ones. Recreating the rpm package on the local machine
is an expensive operation (memory,CPU). If your network connection is
not too slow, you benefit from disabling .delta.rpm.”

[pegasus]

Except that it doesn't save CO2 at all as the CPU has to work harder.

In HPC we learned a decade ago that computation is cheap, it's data transfer that's really expensive. So these days I go around the world looking for possibilities to bring computation closer to data and to minimize data movement. Recreating rpms doesn't seem like something necessary, a simple binary diff would do. And values in rpm db can be updated with an accompanying script.

[cjcox]

They did it years ago as Fedora hasn't offered them by default for a long time already. It's actually a good decision except for the poor users of the metered Internet.

IMHO, it's somewhat of a very privileged assumption when we think "everyone has 100mbit" or somesuch. And many, even if they have 5mbit or greater, are capped as well. Developers: Made sure my code requires 32G and 12 cores, because it's what everyone has.