Define Letter (A to G) Based Energy Ratings for a Given Software Application

[darren72]

Wouldn't it be nice if software also had an "energy rating". This energy rating is commonplace for household appliances, cars and even houses. Why not software? Assigning a letter, such as A for the best rated and say G for the worst rated is easily understood.

If we were able to provide guidance as to which rating should be applied for a given application based on the SCI it might motivate the industry to move towards green software more readily. e.g. a score of 60-70 is A, 70-80 is B and so on.

Therefore, my question would be, would there be a means to translate the SCI into a letter-based score system?

[atg-abhishek]

Thank you @darren72 for sharing that - what was the outcome of the discussion with SWG in the previous meeting? Would this be something that could emerge from the work on the reporting standards? Would we need to work with the SCI Open Data group / think about what the industry benchmarking might look like to make a determination on how to create the different slices? cc @Henry-WattTime @jawache

[jawache]

This comes up a lot, so much so I think we/i need an official position piece about it 😃

The main challenge is that software is not a fridge, an objective rating is possible for a fridge but for a software solution it becomes a lot more subjective.

A) It's easy to know the boundary of a fridge just by looking at it, with software it's a lot more subjective and very hard to verify. We've got some language in the spec but it's very limited.

B) Emissions depend on a number of factors, e.g. the 'ls' command is going to be incredibly efficient and perhaps scores an A, how would you score windows explorer? A D? But it does a lot more. How would you solve this objectively without bringing a human opinion into it.

I think a more reasonable approach would be to have a scorecard for processes and features.

Do you have a process to monitor the SCI score of your application in production?
Has your team completed the green software training?

If you answer yes to 100% you get an A, 90% you get a B etc...

IMO we should be scoring the "development process" not "a specific version of an application", ls and windows explorer can both score A, even if they have wildly different feature sets as long as they have strong process's in place to ensure all future versions of the software emit the least carbon possible.

My 2c!

[Henry-WattTime]

WG: to explain current position that the spec is not mature enough for comparability and star ratings.
Asim, it comes up a lot, so people care about it, means we should consider it carefully
Current intent of spec is to catalyze action, thought piece to follow,
@jawache to coordinate blog post in the queue, in months timeline
In interim should probably be somewhere in the guidance.

[atg-abhishek]

@Henry-WattTime @jawache did we end up publishing on this?

[jawache]

No we didn't 😕 @NAMRATA-WOKE and @ursvill let's bring this up in our content meeting next week.

[NAMRATA-WOKE]

@jawache made a note. Thanks.

[markus-gsf-seidl]

While the Author might have A++++, A+++, A++, A+, A to F system of the EU for laptops and other hardware in mind, I was reminded of the Nutri-Score. I see them when going shopping where healthy food (ex cheese) has now a D rating and highly processed food (energy drinks, protein shakes) do have an A rating.
I guess if we introduce a very simple star or letter label, than the industry will try to game it without reducing CO2.
So this will be a very difficult thing to get right, if not impossible.

I can understand the comparison and competition part this is coming from, but the competition already lies in the score: If SCI > 0 => you have work to do.

Initially I was thinking about an Ranking system: Rank all the "same" software and assign them numbers, like a leaderboard. But then everyone would try to get it's own definition of "same" or misreport features to have it's own list and be at rank "1".
I personally think that the SCI is already good enough and if something wants to be reported, report the eqCO2/Request directly and provide details about how you calculate this.