summaryrefslogtreecommitdiff
path: root/issues
diff options
context:
space:
mode:
authorFrederick Muriuki Muriithi2024-01-29 04:12:10 +0300
committerFrederick Muriuki Muriithi2024-01-29 04:15:36 +0300
commit51637eda561b24b7afd9e1f0b9638652ae33c20d (patch)
treec545a599e01f5f38b56a7f0789a65d21565c05dc /issues
parenta9819e2880cedfff736c3acf2a4003fdf5a691bd (diff)
downloadgn-gemtext-51637eda561b24b7afd9e1f0b9638652ae33c20d.tar.gz
New issue: Error Handling.
Diffstat (limited to 'issues')
-rw-r--r--issues/error-handling-external-errors.gmi29
1 files changed, 29 insertions, 0 deletions
diff --git a/issues/error-handling-external-errors.gmi b/issues/error-handling-external-errors.gmi
new file mode 100644
index 0000000..9208a5d
--- /dev/null
+++ b/issues/error-handling-external-errors.gmi
@@ -0,0 +1,29 @@
+# Error Handling: External Errors
+
+## Tags
+
+* assigned: fredm
+* status: open
+* type: bug
+* priority: high
+* keywords: error handling
+
+## Description
+
+We have been consistently moving a lot of our features out of GN2 to external API services like GN3 and GN-Auth. This means that when we have errors in the external service(s), then GN2 does not have direct access to the error(s) to handle it.
+
+So far, the error-handling for such services has been inconsistent at best. In this issue, we propose that such errors be handled by simply raising an exception with the passed in error information, to wit:
+
+> The better approach may be to simply raise this as an exception so that some
+> exception handler somewhere down the line can handle it.
+
+> Here's the larger philosophy. The exception system of a programming language
+> is meant to handle errors, and is well-suited to this purpose. But, we use
+> different backends connected over an API and break the exception system. Our
+> goal must be to bridge the API gap so that exceptions can cross over. Thus,
+> we enable the exception system to do the job it was always meant to do.
+
+> - Arun Isaac
+> => https://github.com/genenetwork/genenetwork2/pull/830
+
+We can then maybe incorporate the use of UUIDs in the errors, to help with tracking the errors in the logs where necessary.