1 |
On Sun, 22 Mar 2020 15:11:23 +0100 |
2 |
"Haelwenn (lanodan) Monnier" <contact@×××××××××.me> wrote: |
3 |
|
4 |
> I think it might be better to fix bugzilla to be able to send multiple attachments at once. AWS S3 might be okay for the long term but I've often seen paste services being used and most of them expire in a week/month, so you can easily loose the content before it can be read or fixed and absolutely no hope to have it be readable if it's an old bug that might have appeared again. |
5 |
|
6 |
Part of my objections to using Bugzilla for this is due to the |
7 |
attachments frequently being unuseful. |
8 |
|
9 |
Not that *all* of them are useless, but due to the pattern of |
10 |
automation, no sentience evaluates if a given log is useful or not, so |
11 |
there are usually many more to wade through that don't help. |
12 |
|
13 |
Add to this the whole "bugzilla limits attachment sizes", and the |
14 |
workaround being "upload compressed versions which are harder to read", |
15 |
this creates a lot of attachments that can't even be read directly in |
16 |
your browser (even though they're just text, the size demands |
17 |
decompression first). |
18 |
|
19 |
And due to the sizes involved, I worry about the burden it places on |
20 |
the database to perform this at scale. |
21 |
|
22 |
IMO, Databases are just not the tool to use for binary data storage. ( |
23 |
Though I suspect perhaps bugzilla has an out-of-database binary file |
24 |
storage thing? idk ). |
25 |
|
26 |
It feels like that mistake people make when they commit large binary |
27 |
assets into git, where they really want something out-of-band more |
28 |
optimized for it. |
29 |
|
30 |
And like git, once its submitted, there's no "delete" as far as I know, |
31 |
... which, given the ratio of useful data vs the size, seems silly. |
32 |
|
33 |
So the net result is a lot of downsides due to silly limitations in the |
34 |
choice of platform. |
35 |
|
36 |
I don't want to discourage automated testing, but I do want to champion |
37 |
a better platform to store this data (and more importantly, aggregate |
38 |
this data). |
39 |
|
40 |
Just so its clear, my dream is a system like CPAN Testers, where |
41 |
anybody can submit results regardless of who they are, with no quality |
42 |
requirements on the data submissions, ... without necessitating that |
43 |
each and every failure be communicated to a bug report. |
44 |
|
45 |
That way, the 3rd party platform collects information that indicates |
46 |
where a bug *could* be, and once an actual bug is found in this |
47 |
information set, the bug is opened, citing the relevant details. |