Open mcdonnelldean opened 8 years ago
Just to clarify everybody else, it's probably worth explaining why. Il giorno mer 10 feb 2016 alle 22:59 Dean McDonnell < notifications@github.com> ha scritto:
Tentacoli now supports browser and will enable live and historic requests that are pattern matched. We will move to this over nes.
https://github.com/mcollina/tentacoli
cc/ @mcollina https://github.com/mcollina
— Reply to this email directly or view it on GitHub https://github.com/vidi-insights/vidi-dashboard/issues/123.
@mcollina Good point, I thought, because it is awesome would do :D
Will edit above.
@mcollina Question on this, based on the first cut I wrote I need some way to Id a client. Is this something Tentacoli can do or am I best just sending a nid from the client in the request?
Each request in Tentacoli gets a uuid. Should we look at ways to surface this? https://github.com/mcollina/tentacoli/blob/master/tentacoli.js#L238
@mcollina I think I need per client. A client might have multiple subscriptions running.
I think you will have to do it on top of tentacoli, as par of the websocket-stream/hapi/authentication stuff.
@mcollina Agreed, I have to write a client-side wrap anyway to add context stuff so it makes sense to do it there
I will put all of it in an hapi plugin, similar to nes.
@mcollina Great idea, I will make this it's own issue on Tentacoli
Tentacoli now supports browser and will enable live and historic requests that are pattern matched. We will move to this over nes.
https://github.com/mcollina/tentacoli
Reasoning
Vidi works on a system of metrics. These metrics are requested like so,
/group/metric
and are generally very loose in what they return.One of the features we require of a metric query layer is that it can return both live and historic data; streams make perfect sense for this. We also have the issue whereby different screens will look for the same data but in different shapes or contexts.
Tentacoli is a stream multiplexer that allows the above. You can essentially request / response to the server once per metric, and have it return a unique stream of data. Since each request can have multiple streams we can also batch a whole screens data to one call but still maintain individual response streams.
Nes can't do that :D