Serialisation of virtual threads

[Ron Pressler]

Hi.

Scheduling points were never explicit at any iteration of the project, and the decision to
use the java.lang.Thread API rather than some new class should not affect the serialization
feature one way or another. It is, however, an advanced feature, and we’d like to stabilise
the more basic features before we get around to the advanced ones. In Quasar, too 
serialization came later.

— Ron


On 1 July 2020 at 12:27:03, Dimos Raptis (dimos.raptis at r3.com) wrote:

Hello everyone,  

I remember the initial Loom proposal mentioned serialization of fibers and I can also see that being mentioned under the "Missing Features" section of the home page.  
After reading the "State of Loom" post, I noticed that fibers are being replaced by "virtual" threads and will comply with most of the existing APIs. It also seems scheduling/suspension points are not explicit anymore and they become to some extent an implementation aspect of the feature, not necessarily visible to the user.  

I was wondering if that has any implications on the ability/desire to implement serialization of them? Could you share any plans on this area?  

We are making use of Quasar fibers at the moment and we would like to migrate to the native API. But, in order to do this we would need the ability to suspend-serialize and deserialize-schedule these virtual threads at specific points.  

Thanks and apologies if this has been answered already,  
Dimos