Watch the video, transcript and sources
https://www.youtube.com/watch?v=7b1i5wE_sjQ
+ Compatible servers can create an exit option
- Existing workflows still need a migration rehearsal
+ Running the software brings operating responsibility homeSalesforce advertises its Pro suite at a hundred dollars per user per month, billed annually. One more person means one more seat on that plan. OccupySaaS wants companies to have another choice: run compatible software on infrastructure they control.
Its Salesforce project, SalesCompel, is early. There is real persistent customer-record code in the repository, alongside a separate exploration prototype that keeps its edits in memory. An exit plan has to account for both the working slice and the unfinished product around it.
Keep the application talking
The basic idea is understandable. An application already sends requests through a software client. A compatible server can accept those requests at another destination, preserving more of the application the company has already built. That reduces one source of migration work.
The customer's setup still needs checking. Records, custom fields and permissions have to retain their meaning. A workflow that depends on an unsupported operation belongs on the acceptance list before anyone calls the migration complete.
OccupySaaS also develops ZeroAuthy, an Auth0-compatible identity server. Its matrix records checks with named client libraries and versions, including the React client. Some behaviors have client checks, while others have tests of the server alone. That distinction makes the evidence more useful than a broad compatibility slogan.
The OpenID Foundation's own list includes a particular ZeroAuthy commit for three provider profiles. This is a defined conformance self-certification scope. Multi-factor sign-in and an operational hardening sweep remain on the project's later roadmap. The listing gives a prospective operator one bounded claim to inspect; a production decision needs broader evidence about the deployment and the workflows it will support.
Which SalesCompel screen are you looking at?
The Go/PostgreSQL candidate stores Account and Contact records and supports a limited set of operations and queries. Its implementation guide documents the served surface and the exclusions.
The separate Vue exploration prototype uses synthetic data. Its edits remain in memory and do not call the backend. A screen can provide useful feedback about a proposed product while its migration and wider workflow support are still being built.
The builders describe the unfinished work in public. No full program milestone has a recorded exit-gate pass, and production certification has not been reached. Their remaining estimate covers hundreds of bounded work packages. It is a scope estimate, with no delivery date implied by the number.
The website uses layoff and protest language. We have not verified the contributors' employment histories. Its timeline dates the public projects to mid-October; today's event is the fresh discussion. No newly established launch is being claimed here.
The maintenance bill follows you
On October 21, PostgreSQL committer Robert Haas described AI-assisted bug discovery outpacing human review. He linked that pressure to late feature reversals and the version nineteen delay. Models can draft fixes, while humans still check correctness and project standards. His dated report illustrates an operating obligation that open-source software also carries.
In today's discussion, HN commenter ozim points to backups and security updates among the work customers buy. A self-hosting company needs an owner for those jobs. Removing the software's per-seat license charge does not supply an operating team.
The off switch is a specific job
SalesCompel's early backend guide makes the responsibility concrete. To revoke a provisioned token, an authorized database operator deletes its stored digest row. The digest is the token's fingerprint in the database. The guide places fuller token lifecycle and public login in future work.
When an employee leaves, their access needs to stop. The early candidate gives the operator a documented step. A migration decision has to include who performs it and how the company verifies the outcome.
The clean-room policy also deserves a read. It excludes confidential employer material and acknowledges contractual risk around competitive API observation. Apache licensing gives permission to use this code within its terms; the operator's legal review remains a separate job.
Verdict: NEEDS REVIEW
I'd test the supported slices because a compatible open server can create a useful exit option. I'd gate production on migration evidence, recovery checks and clear operating responsibility. The public roadmap gives a buyer something concrete to interrogate before moving customers behind it.
Would you self-host your company's CRM, or pay somebody to operate it?
Sources
Today's Hacker News discussion and ozim's operating-work comment
SalesCompel early backend guide: implemented surface and token revocation
ZeroAuthy README and roadmap, compatibility matrix and clean-room policy
OpenID Foundation listing and definition of self-certification
Robert Haas, October 21: Postgres 19 is late, and it's AI's fault
Cover portrait: Christie Hemm Klok / Salesforce via Wikimedia Commons, CC BY-SA 2.0. Cropped/background removed; cover adaptations use the same license.
And that's the diff for today. I'm Niko from Axrisi. Merge responsibly.