> I hate TAGG because it's built around the lie 
> <https://groups.google.com/d/msg/nodejs/ELsE4-L_oxM/HHqEuyp7HHIJ>that 
> zero-copy data sharing would be at all possible. V8 just isn't built to do 
> it, and making it so would be a monumental task. Google may be able to pull 
> it off it they cared, but the core team can't. Jorge and Bruno don't even 
> know where to look. Call me again when you own a billion dollar company. Or 
> when you're reborn as a crazy Russian super 
> hacker<https://github.com/sheremetyev/w16>. 
> Zero-copy Buffer/TypedArray sharing *may* be possible but copying binary 
> data is even less likely (by at least an order of magniture I'd say) to be 
> the bottleneck in your app than serialization.
>

Multi-threading, even without zero copy, would allow for faster context 
switching as well. Context switching is bad, process-per-thread is good and 
helps avoids context switching, but there is a gain even if there is not an 
easy zero copy benefit to be had.

Not clamoring for this, zero-copy is far more my interest, but it is a gain 
for places where concurrency has to happen outside of any given Node 
process.
 

> In the meantime, ToString'ing a function and eval'ing it on another 
> thread, are you kidding me? Heavy computation never takes more than a 
> single function right?
>

Easy Bert, let's not trash something that works & does more for some 
people. It doesn't have to appeal to everyone to be a good thing.
 

> Isolates was our (or rather, Ben and Ryan's) attempt to do whatever TAGG 
> might achieve in a proper way, but in the end we figured out that it has no 
> benefits over using child processes. Whatever TAGG does, you can trivially 
> achieve the same with child processes. Ryan even wrote an example 
> fibonacci server <https://gist.github.com/2018811> to demonstrate this. 
> Obviously nobody cared because this is entirely irrelevant.
>

At the time, the only explanation we got for it being dropped was that it 
was 15 words, and those rang to a different tune than "had no benefit":

ultimately turned out to cause too much instability in node's internal 
> functionality to justify continuing

https://groups.google.com/d/topic/nodejs/zLzuo292hX0/discussion 

I still am very much interested in technical discussion around this 
feature, know it abstractly but not deeply, and the cause of it being 
shelved remains shrouded in mystery. I'm sorry we all feel so dragged into 
the mud every time it gets raised, thank you for helping share more views 
which further enlightenment us: any willing to chip in technically on this, 
thanks (Ben! Jorge! Bert!).

-r

-- 
Job Board: http://jobs.nodejs.org/
Posting guidelines: 
https://github.com/joyent/node/wiki/Mailing-List-Posting-Guidelines
You received this message because you are subscribed to the Google
Groups "nodejs" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to
[email protected]
For more options, visit this group at
http://groups.google.com/group/nodejs?hl=en?hl=en

Reply via email to