[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [p2-dev] A scary thought
- From: John Arthorne <arthorne.eclipse@xxxxxxxxx>
- Date: Wed, 9 Jun 2010 09:59:25 -0400
- Delivered-to: email@example.com
- Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:in-reply-to:references:from:date:message-id:subject:to :content-type; b=t3FGGci0gUjNE2YhHudQN3Vq+AzQdcqtXrQP2bWOACvU6IO5uD0XERZw92OlOsxruf QAKoOUPna6AqBLNSKQjI09BOCQSue1YZsRhl/eFVwqoQe08V/w+oP7mGoWj4yol+ihol cygsjfDT3ojwYdtTBgmGGxloPGX39tB+JS444=
Luckily that's not how it works. The mirror itself is never consulted as a p2 repository, i.e., we never read the artifacts.jar/compositeArtifacts.jar at the mirror location. When we have a mirror location, we just replace the prefix in the artifact URL with the mirror location, and directly fetch the artifact over HTTP. The mirror must contain the same artifacts in the same disk layout as the repository that it is mirroring. If it is a different type of repository it won't work (the artifact won't be found and it will try again on the next mirror).
On Wed, Jun 9, 2010 at 5:08 AM, Thomas Hallgren <thomas@xxxxxxx>
I've had some discussions regarding how to set the mirrorsURL. One idea that came up was to let it point to a composite. If I'm correct that is a very bad idea because it means that in order to find an artifact, p2 will do the following:
1. It finds an artifact composite. The composite appoints an artifact repo.
2. The artifact repo contains a mirrorsURL. p2 picks that up.
3. p2 selects a mirror.
4. Now we're back at #1.
It will take some time to find that artifact...
Is this the way it works? If it is, perhaps this needs to be communicated.
p2-dev mailing list