On doctors and poking. On doctors and poking.

So there's this traditional dialog, it goes something like this:

Patient: Doctor, doctor, it hurts when I do that....
Doctor: ... then don't do that.

In a closed source world this is pretty much the only version of the problem.

What happens is that the Doctor, being paid (monetarily and dignitarily) dispenses some advice. At a minimum: Don't do that. But possibly: why not do <something else instead>?

Now, things get interesting in an open source environment. Actually it's happening with modern medicine too.

Instead of a patient saying to the doctor: It hurts when I do <this>. The patient asks: How can --

Hrm, this is tough. let's look at a better example of the doctor problem:

Patient: Doctor, doctor, i turn red when i'm out in the sun too long.
Doctor: then don't stay out in the sun without sun screen.

Our Patient (has alergies to face paint): Doctor, doctor, i need help applying this white face paint.
Doctor: puzzled. Why are you applying face paint?
Patient: Because whenever I'm out in the sun too long my face turns red.
Doctor: But you're alergic to face paint. And it doesn't block the UV radiation which causes skin cancer.
Patient: oh.
Doctor (to self): why didn't the patient ask about the sunburn?

How does this manifest in the software world?

I'll provide examples occasionally:

• User hooks up multiple mozillas to a bookmarks file
• User observes that when a mozilla quits, it clobbers the bookmarks file
• User asks: how do i make mozilla write changes to the bookmarks file immediately
• What the user didn't notice: mozilla doesn't check to see if the bookmarks file is changed, so just updating the file more frequently doesn't solve the actual problem (<describe problem>).

what moral would i like people to learn from this?

Instead of asking for help trying to do something which people think would fix the problem they're having, simply describe the problem and ask what can/should be done.

A better way to pin DNS entries without suffering from cache poisoning. A better way to pin DNS entries without suffering from cache poisoning. If the pinned entry is unavailable and the user wants to contact the site, provide a dialog:

+------------------------------------------------------------+
| Host unavailable: 66.66.66.66 "foo.evil.com"               |
+------------------------------------------------------------+
| -=/ =- The server you are trying to reach is unavailable.  |
|                                                            |
|        Host: foo.evil.com                                  |
|        IP:   66.66.66.66   Protocol: HTTP    Port: 80      | 
|                                                            |
|        The following addresses are listed as alternatives. |
|        If you trust the addresses, you may choose to tie   |
|        them to [66.66.66.66,foo.evil.com].  Otherwise you  |
|        may connect to them and prevent them from accessing |
|        the old entries.                                    |
|                                                            |
|        [x] Do not tie the alternatives to 66.66.66.66      |
|        Alternatives: |[     10.0.0.1]|^|                   |
|                      |[  192.168.0.1]| |                   |
|                      |[  192.168.0.2]| |                   |
|                      |[    127.0.0.1]|v|                   |
|        [ Connect ] [ Replace ] [ Stop ]              ([?]) |
+------------------------------------------------------------+

If the user chooses to tie the alternatives to 66.66.66.66/foo.evil.com then Mozilla will add the selected entries to the DNS cache for foo.evil.com and allow foo.evil.com under any of those IP addresses to read cache data from the others and connect to each-other at will.

If the user chooses to connect and not tie the alternatives to 66.66.66.66/foo.evil.com then all cache entries for 66.66.66.66/foo.evil.com will be unavailable to the new foo.evil.com, and the DNS cache for foo.evil.com will be changed from 66.66.66.66 to the newly selected entries.

Note: there once was a proposal for doing this tossed by me in an email, I need to dig it up and transcribe any useful details into it.

Note: https shouldn't use pinning at all since the certificate should protect the cache.

What should a user be able to do with an editor?

Use composer to Edit this link: first list

What would a user migrating from netscape 4 composer expect to be able to do?

Use composer to Edit this link: this list

problem
solution
db implementation details
sample ui for the product/component fields

What's wrong with mozilla today? What's wrong with mozilla today?

• It's way too complicated.
• It has way too many preferences.
• It has way too much text which requires localization.
  • The localization tags are a royal pain.
• It reinvents too many things.
• It doesn't integrate with the os.

Doesn't firebird solve that?

No :)

So what could be done?

Well, imagine a browser which had virtually no text in its ui. There's text for error messages and text for preferences, but ideally while browsing you see neither of these.

Note that this pair is a lark. I'm not actively suggesting such a product. It's just an interesting conceptual demonstration of what someone could do.

What should you do when someone asks for a way to watch all bugs in a product? What should you do when someone asks for a way to watch all bugs in a product?

tell them to watch all QA's for the product.

If they don't want to spend the time to do this, then they probably don't have the time to read the mail they think they want. By making them think about the number of components and the amount involved you can encourage them to consider what they're getting into. If they later decide that they've gotten too much mail they will have a way to reduce it by only watching some of the QA's.

If a project lead watching a bunch of QA's is overwhelmed and notices that there's no way to reduce the mail because most of it comes from a single QA, then the lead will hopefully reach the conclusion that the QA's coverage area is too large and some of the work should be given to someone else.

the latest feature request: Being able to watch by product the latest feature request: Being able to watch by product

Executive Summary: not worth implementing

Any particular reason?

Reasons:

It shouldn't be easy for people to kill themselves or annoy the world.
Watching a product is at least one if not both.

Why people think they want it:

They have project teamleads who want to see all bugmail related to a product, but not all the mail of the people involved with said product.

What those people should do instead:

Manually watch all of the qa's.

Why can I make this suggestion?

Speaking as the person who probably gets the most bugmail of anyone at bmo (5.7%, 5773 in one week, reported by justdave on 9/8/2003).

You'll find two problems:

1. The person will be overwhelmed
2. People will get annoyed at seeing this person getting bugmail no matter what they do
   1. If there's a second product in the installation then someone might decide to use it to do their work
      Based on the fact that the request is for per product watching this implies there's more than one product (if there isn't then you can just modify the template to cc this team lead to all mail...)

   2. If there isn't, someone might decide not to do any updates in bugzilla until the bug is finished

It's true that in corporate environments you can theoretically discourage both practices, but you'll find that you can't; someone can easily end up doing all work outside of bugzilla and only posting when they're done.

Is it a cultural thing?

In some ways yes. But really it's a human thing.

Suppose you're a pretty small team (~15 people)

How many qa contacts do you have?
and how many of your qa contacts watch touch multiple products?

Perhaps:

four - one for each main product , + one lead

So why doesn't the lead watch the qa contacts?

It is simple, possible today, and it should fit your needs.

But you hadn't thought of that option?

The problem is that i have :)