version 3.0.0 - "candle clock"
- The catalogue of laws grew to 38, every model on its published formula, and the
package wears a logo of its own.
- Life tables grew: build one from life expectancy, close or extend the open age,
and name the average time lived in the first year - now by the Human Mortality
Database rule, so life expectancy at birth differs slightly from the last release.
- Downloads work again, the models and life tables are now tested almost
completely, and the package speaks up less, as a note rather than a warning.
- The plots are new throughout, in one house style.
- The documentation is new end to end and now lives on a website: a guided
introduction, every model with its published formula, and the life tables.

version 2.13.1 - "elbow room"
- Website figures are taller and sharper, the survivorship panel shows l(x), and
article tables give short columns only the width they need.

version 2.13.0 - "chalk outline"
- The package now speaks up less and more clearly: where it adjusts the data
you gave it, it says so once, as a note rather than a warning, and a value you
supply for the last, open age interval is used as given instead of quietly
replaced.
- Axis tick labels are short and readable (20k, 100k rather than 2e+04 and
1e+05), a panel carries only as many as it has room for, and the figure
subtitle is drawn larger than before.
- Fitted models, life tables and conversions all draw their own figures in one
house style, with the margins around every panel reworked so the labels clear
the data, and any layout you ask for comes out whole.
- The average time lived in the first year now follows the Andreev-Kingkade
rule, and a life table can be built from a life expectancy vector.
- New models for the first years of life, two for old age and a straight-line
law join the catalogue, with every model on its published formula.
- Downloads from the mortality databases work again, and the models and the
life tables are almost fully tested.

version 2.12.1 - "quiet attic"
- Old working files the package no longer needs have been cleared out, and
the readme has been tidied up.

version 2.12.0 - "full picture"
- The documentation is new end to end: a guided introduction, a tour of the
mortality models and their formulas, and a walk through the life table methods.
- The package has a website now, with every function indexed and every article
readable online. The installation guide has moved into the README.

version 2.11.0 - "new clothes"
- The plots are new throughout: a fitted law draws its fit and a residual
diagnosis, a life table its survivorship, hazard, deaths and expectancy, and
a conversion its two curves. The panels split as you like.
- The fit summary grew to match: the fit window, the method, the deviance,
the R-squared and RMSE. A conversion result now plots and prints itself.
- The life table stays quiet when its own ax method assigns the open
interval; only your own values are announced when replaced.

version 2.10.4 - "small print"
- A failed login now says the session was not set up, and the test suite is
half the size and about four times faster.

version 2.10.3 - "hex appeal"
- The package wears a new logo: two curves of a life, with the little figures
between them thinning out as the years pass.

version 2.10.2 - "safety net"
- The models and the life tables are now tested almost completely, so a change
that breaks them is caught before it reaches you.

version 2.10.1 - "open shelf"
- The mortality models can be called by name again, as before, without
crowding the help index.

version 2.10.0 - "late bend"
- Three models are new: two forms of the Gompertz-Makeham family for old age,
and the 1725 straight-line law, which warns that it applies only to the ages
it was fitted to.
- The Weibull model no longer invents a value at birth; the first age is now
reported as missing, which fixes its fit for the first years of life.
- The catalogue now flags which existing models describe the same curve.

version 2.9.1 - "quiet shelf"
- The death models no longer crowd the help pages; the model catalogue now
lists them in one place, with where each fits and its source paper.

version 2.9.0 - "backwards glance"
- Life tables can now be built from a vector of life expectancy. The
function works out the survival curve and the death rates that reproduce the
supplied figures.
- The conversion function accepts life expectancy as a starting point as
well, so it can now convert between any pair of the five table inputs and
the seven table columns.

version 2.8.0 - "second rule"
- Life tables now use the Human Mortality Database rule for the average
time lived in the first year of life, named after its authors, Andreev
and Kingkade. It is more accurate and is now the default.
- Life expectancy for babies therefore changes slightly. The previous
rule is still available.

version 2.7.0 - "long tail"
- Life tables can now be closed more accurately at the open age, either by
using a mortality model to correct the last interval in place or by extending
the table to an older age before closing.
- How long people live on average in the interval where they die is now
chosen by name, and by default it follows the standard unadjusted formula.
Previously an adjustment was applied automatically at the youngest ages
whenever a sex was given, so life expectancy for babies differs slightly
from before.

version 2.6.1 - "second opinion"
- Fit summaries now judge the quality of a fit more reliably, and models
fitted from death counts and exposures are compared on the correct scale.

version 2.6.0 - "early hours"
- Four new models for the first years of life are available: the negative
Gompertz, the Pareto II, the shifted power, and the truncated power model.
- The truncated power model needs age detail finer than single years; with
coarse data it falls back to the shifted power model and says so.

version 2.5.1 - "sign post"
- The help page for the 1870 mortality model now explains why its middle
term is written with a minus sign.

version 2.5.0 - "twin anchor"
- Life tables can now use either of the two published versions of the
childhood mortality adjustment: the one used until now, and the original
1983 rule. The default is unchanged, so results stay the same as before.

version 2.4.0 - "swept deck"
- The internet address shown for the Canadian mortality database now works;
it used to point to a page that could not be opened.
- Four numbers in the country list were missing their commas, and a few names
were misspelled in the help pages.
- Installing the package now needs one fewer outside library.

version 2.3.1 - "open channel"
- Data downloads from the Human Mortality Database work again; the login had
stopped working, so downloads with credentials always reported a failure.

version 2.3.0 - "steady harbor"
- Comparing models and building life tables is more reliable now, most
noticeably for babies and at the oldest ages.
- The affected models now follow their original published formulas, and
one of them has a name of its own.
- One old model was added. Error messages, plots and help pages improved.

version 2.2.2
- Documentation update for testing functions;

version 2.2.0
- Simplified optimisation process by using PORT routines for all models. 
This should improve the fitting process and the convergence in some cases;
- Documentation update for main functions;

version 2.1.2
- update web source for Australian Human Mortality Database;

version 2.1.1
- provide anchors to all documented external \links{};

version 2.1.0
- replace several stop errors with messages so the failure to connect to databases (for
whatever reason) results in a graceful ending of the process and to avoid triggering
CRAN random error;

version 2.0.3
- Change interaction with external resources and update documentation;

version 1.9.9
- fix hmd links and add more messages in case of failure to connect to external
sources;

version 1.9.3
- fix minor bugs, links and dois

version 1.8.5
- make the function using Internet resources to fail gracefully with an
informative message if the resource is not available or has changed.

version 1.8.1
- LifeTable(): fix minor bug that prevented the closing of the life table with
a qx = 1 in some situations;
- Update internal unit tests.

version 1.8.0
- ReadCHMD(): New function useful to download the Canadian Human Mortality Database;
- ReadAHMD(): New function useful to download the Australian Human Mortality Database;
- ReadJMD(): New function useful to download the Japanese Mortality Database;
- convertFx(): improve the speed of the function > 5x;
- availableHMD(): improve the speed of the function  > 100x;
- LifeTable(): the argument 'ax' now accepts as input a numerical vector as well
allowing the user to decide the timing of occurrence of death in each age interval;
- Update LICENSE: MIT. Drop GPL-3.

version 1.7.0
- New models: Kannisto-Makeham (1998) and Strehler-Mildvan (1960);
- ReadHMD() new functionalities. The function can now download cohort life tables,
cohort life expectancy and exposure-to-risk by Lexis triangles data.

version 1.6.0
- convertFx(): New function! Easy conversion between the life table indicators;
- MortalityLaw(): Remove scale.x argument. Now, the function will identify
by default which laws require a scaled x in fitting;
- MortalityLaw(): When a mortality law is fitted on a subset of the input data,
using fit.this.x argument, the residuals and fitted values are computed and plotted
for the entire age range (in input). Previously only the fitted values followed this
approach;
- MortalityLaw(): set the argument `show = FALSE` as default;
- Drop `diagram` suggestion from DESCRIPTION and vignettes.

version 1.5.0
- LawTable(): Compute life tables from parameters of a mortality law;
- Fix bug in the ReadHMD() function due to HMD migration from http to https.

version 1.4.0
- Compute e[x] at the beginning of the age interval in LifeTable(). This affects only
the ex column in the life table, all the other indicator remain unchanged;
- New mortality model: Gamma-Gompertz (Vaupel.1979 version);
- Fix bug related to fit.this.x argument in MortalityLaw();
- More rigorous referencing in availableLaws() function.

version 1.3.0
- Add generic functions: deviance, df.residual;
- Update vignette.

version 1.2.0
- Vignettes;
- Fix predict.MortalityLaw bug for cases where the input is of class matrix;
- Solution for 'Kostaki' inflexion point due to high correlation between estimated
parameters;
- Add 'scale.x' argument in MortalityLaw function;
- Add 'ax' argument in LifeTable function;
- Add 'show' argument in ReadHMD function;
- Other minor changes.

version 1.0.5
- Fix issues with minpack.lm not being imported.

version 1.0.0
- First stable version (on CRAN).
